IBM NeXtScale System, NeXtScale n1200, NeXtScale nx360 M4 Planning And Implementation Manual

ibm.com/redbooks
IBM NeXtScale System Planning and Implementation Guide
David Watts
Jordi Caubet
Duncan Furniss
David Latino
Covers the n1200 Enclosure and nx360 M4 Compute Node
Addresses power, cooling, racking, and management
Front cover
IBM NeXtScale System Planning and Implementation Guide
July 2014
International Technical Support Organization
SG24-8152-01
© Copyright International Business Machines Corporation 2013, 2014. All rights reserved.
Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP Schedule Contract with IBM Corp.
Second Edition (July 2014)
This edition applies to:
򐂰 IBM NeXtScale n1200 Enclosure, machine type 5456 򐂰 IBM NeXtScale nx360 M4, machine type 5455
Note: Before using this information and the product it supports, read the information in “Notices” on page vii.
© Copyright IBM Corp. 2013, 2014. All rights reserved. iii
Contents
Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Now you can become a published author, too! . . . . . . . . . . . . . . . . . . . . . . . . xii
Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
Stay connected to IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Summary of changes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
July 2014 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
April 2014, Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Chapter 1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Evolution of data centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Scale out applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Executive Summary of IBM NeXtScale System . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 IBM NeXtScale n1200 Enclosure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 IBM NeXtScale nx360 M4 compute node . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Design points of the system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 This book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2. Positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Market positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Three key messages with NeXtScale . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Optimized for workloads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 IBM System x overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 NeXtScale System versus iDataPlex . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 NeXtScale System versus Flex System . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 NeXtScale System versus rack-mounted servers . . . . . . . . . . . . . . . . . . . 20
2.6 Ordering and fulfillment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Chapter 3. IBM NeXtScale n1200 Enclosure . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.1 Front components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.1.2 Rear components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.3 Fault tolerance features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Standard chassis models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
iv IBM NeXtScale System Planning and Implementation Guide
3.3 Supported compute nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Power supplies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Fan modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Midplane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.7 Fan and Power Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.7.1 Ports and connectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.7.2 Internal USB memory key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.7.3 Overview of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7.4 Web GUI interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.8 Power management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.8.1 Power Restore policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.8.2 Power capping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.8.3 Power supply redundancy modes . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.8.4 Power supply oversubscription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.8.5 Acoustic mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.9 Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.9.1 Physical specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.9.2 Supported environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Chapter 4. IBM NeXtScale nx360 M4 compute node . . . . . . . . . . . . . . . . . 59
4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.1 Physical design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 System architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Specificiations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.4 Standard models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.5 Processor options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.6 Memory options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6.1 DIMM installation order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Internal disk storage options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.7.1 Controllers for internal storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.7.2 Using the ServeRAID C100 with 1.8-inch SSDs . . . . . . . . . . . . . . . . 84
4.7.3 HDDs and SDDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.8 IBM NeXtScale Storage Native Expansion Tray . . . . . . . . . . . . . . . . . . . . 86
4.9 IBM NeXtScale PCIe Native Expansion Tray . . . . . . . . . . . . . . . . . . . . . . 90
4.10 GPU and coprocessor adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.11 Embedded 1 Gb Ethernet controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.12 PCI Express I/O adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.12.1 Mezzanine adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.12.2 Single-slot riser card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.12.3 Network adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.12.4 Storage host bus adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.13 Integrated virtualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.14 Local server management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Contents v
4.15 Remote server management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.16 External disk storage expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.17 Physical specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.18 Operating systems support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Chapter 5. Rack planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.1 Power planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.1.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.1.2 PDUs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.1.3 UPS units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.2 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.3 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.4 Racks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.4.1 The IBM 42U 1100mm Enterprise V2 Dynamic Rack . . . . . . . . . . . 125
5.4.2 Installing NeXtScale System in other racks . . . . . . . . . . . . . . . . . . 131
5.4.3 Rack options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.5 Rear Door Heat Exchanger. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.6 Top of rack switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.6.1 Ethernet switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.6.2 InfiniBand switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.6.3 Fibre Channel switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.7 Rack-level networking: Sample configurations . . . . . . . . . . . . . . . . . . . . 146
5.7.1 Non-blocking InfiniBand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.7.2 50% blocking InfiniBand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.7.3 10 Gb Ethernet, one port per node . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.7.4 10 Gb Ethernet, two ports per node . . . . . . . . . . . . . . . . . . . . . . . . 150
5.7.5 Management network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Chapter 6. Factory integration and testing . . . . . . . . . . . . . . . . . . . . . . . . 153
6.1 IBM standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.2 Testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.3 Documentation that is provided. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.3.1 HPLinpack testing results: Supplied on request . . . . . . . . . . . . . . . 157
Chapter 7. Hardware management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.1 Managing compute nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7.1.1 Integrated Management Module II . . . . . . . . . . . . . . . . . . . . . . . . . 160
7.1.2 Unified Extendible Firmware Interface . . . . . . . . . . . . . . . . . . . . . . 164
7.1.3 ASU. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
7.1.4 Firmware upgrade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
7.2 Managing the chassis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7.2.1 FPC web browser interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
7.2.2 FPC IPMI interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
7.3 ServeRAID C100 drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
vi IBM NeXtScale System Planning and Implementation Guide
7.4 VMware vSphere Hypervisor (ESXi) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Chapter 8. Software stack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
8.1 eXtreme Cloud Administration Toolkit (xCAT). . . . . . . . . . . . . . . . . . . . . 214
8.2 IBM Platform Cluster Manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
8.3 IBM General Parallel File System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.3.1 IBM GPFS FPO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
8.3.2 IBM System x GPFS Storage Server . . . . . . . . . . . . . . . . . . . . . . . 224
8.4 IBM Platform LSF family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.5 IBM Platform HPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
8.6 IBM Platform Symphony family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
8.7 IBM Parallel Environment for x86 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
8.7.1 IBM Parallel Environment Runtime for x86 . . . . . . . . . . . . . . . . . . . 234
8.7.2 IBM Parallel Environment Developer Edition for x86 . . . . . . . . . . . 236
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Help from IBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
© Copyright IBM Corp. 2013, 2014. All rights reserved. vii
Notices
This information was developed for products and services offered in the U.S.A. IBM may not offer the products, services, or features discussed in this document in other countries. Consult your
local IBM representative for information on the products and services currently available in your area. Any reference to an IBM product, program, or service is not intended to state or imply that only that IBM product, program, or service may be used. Any functionally equivalent product, program, or service that does not infringe any IBM intellectual property right may be used instead. However, it is the user's responsibility to evaluate and verify the operation of any non-IBM product, program, or service.
IBM may have patents or pending patent applications covering subject matter described in this document. The furnishing of this document does not grant you any license to these patents. You can send license inquiries, in writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive, Armonk, NY 10504-1785 U.S.A.
The following paragraph does not apply to the United Kingdom or any other country where such provisions are inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION
PROVIDES THIS PUBLICATION “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer of express or implied warranties in certain transactions, therefore, this statement may not apply to you.
This information could include technical inaccuracies or typographical errors. Changes are periodically made to the information herein; these changes will be incorporated in new editions of the publication. IBM may make improvements and/or changes in the product(s) and/or the program(s) described in this publication at any time without notice.
Any references in this information to non-IBM websites are provided for convenience only and do not in any manner serve as an endorsement of those websites. The materials at those websites are not part of the materials for this IBM product and use of those websites is at your own risk.
IBM may use or distribute any of the information you supply in any way it believes appropriate without incurring any obligation to you.
Any performance data contained herein was determined in a controlled environment. Therefore, the results obtained in other operating environments may vary significantly. Some measurements may have been made on development-level systems and there is no guarantee that these measurements will be the same on generally available systems. Furthermore, some measurements may have been estimated through extrapolation. Actual results may vary. Users of this document should verify the applicable data for their specific environment.
Information concerning non-IBM products was obtained from the suppliers of those products, their published announcements or other publicly available sources. IBM has not tested those products and cannot confirm the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products.
This information contains examples of data and reports used in daily business operations. To illustrate them as completely as possible, the examples include the names of individuals, companies, brands, and products. All of these names are fictitious and any similarity to the names and addresses used by an actual business enterprise is entirely coincidental.
COPYRIGHT LICENSE: This information contains sample application programs in source language, which illustrate programming techniques on various operating platforms. You may copy, modify, and distribute these sample programs in any form without payment to IBM, for the purposes of developing, using, marketing or distributing application programs conforming to the application programming interface for the operating platform for which the sample programs are written. These examples have not been thoroughly tested under all conditions. IBM, therefore, cannot guarantee or imply reliability, serviceability, or function of these programs. You may copy, modify, and distribute these sample programs in any form without payment to IBM for the purposes of developing, using, marketing, or distributing application programs conforming to IBM's application programming interfaces.
viii IBM NeXtScale System Planning and Implementation Guide
Trademarks
IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. These and other IBM trademarked terms are marked on their first occurrence in this information with the appropriate symbol (® or ™), indicating US registered or common law trademarks owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries. A current list of IBM trademarks is available on the Web at http://www.ibm.com/legal/copytrade.shtml
The following terms are trademarks of the International Business Machines Corporation in the United States, other countries, or both:
AIX® BladeCenter® Blue Gene® GPFS™ IBM® IBM Flex System™ iDataPlex® Intelligent Cluster™
LoadLeveler® LSF® POWER® Power Systems™ PowerPC® PureFlex™ RackSwitch™ Redbooks®
Redpaper™ Redbooks (logo) ® ServerProven® Symphony® System Storage® System x®
The following terms are trademarks of other companies: Evolution, and Kenexa device are trademarks or registered trademarks of Kenexa, an IBM Company. Intel, Intel Xeon, Intel logo, Intel Inside logo, and Intel Centrino logo are trademarks or registered trademarks
of Intel Corporation or its subsidiaries in the United States and other countries. Linux is a trademark of Linus Torvalds in the United States, other countries, or both. Microsoft, Windows, and the Windows logo are trademarks of Microsoft Corporation in the United States,
other countries, or both. UNIX is a registered trademark of The Open Group in the United States and other countries. Other company, product, or service names may be trademarks or service marks of others.
© Copyright IBM Corp. 2013, 2014. All rights reserved. ix
Preface
IBM® NeXtScale System is a new, dense offering from IBM. It based on our experience with IBM iDataPlex® and IBM BladeCenter® with a tight focus on emerging and future client requirements. The IBM NeXtScale n1200 Enclosure and IBM NeXtScale nx360 M4 Compute Node are designed to optimize density and performance within typical data center infrastructure limits.
The 6U NeXtScale n1200 Enclosure fits in a standard 19-inch rack and up to 12 compute nodes can be installed into the enclosure. With more computing power per watt and the latest Intel Xeon processors, you can reduce costs while maintaining speed and availability.
This IBM Redbooks® publication is for customers who want to understand and implement an IBM NeXtScale System solution. It introduces the offering and the innovations in its design, outlines its benefits, and positions it with other IBM x86 servers. The book provides details about NeXtScale System components and the supported options. It also provides rack and power planning considerations and describes the ways that you can manage the system.
Authors
This book was produced by a team of specialists from around the world working at the IBM Taiwan Systems & Technology Lab (TSTL) in Taipei, Taiwan.
David Watts is a Consulting IT Specialist at the IBM ITSO Center in Raleigh. He manages residencies and produces IBM Redbooks® publications about hardware and software topics that are related to IBM Flex System™, IBM System x®, and BladeCenter servers and associated client platforms. He authored over 250 books, papers, and Product Guides. He holds a Bachelor of Engineering degree from the University of Queensland (Australia), and worked for IBM in the United States and Australia since
1989. David is an IBM Certified IT Specialist, and a member of the IT Specialist Certification Review Board.
x IBM NeXtScale System Planning and Implementation Guide
Jordi Caubet is an IT Specialist with IBM in Spain. He has
seven years of experience with IBM and several years of experience in high-performance computing (HPC), ranging from systems design and development to systems support. He holds a degree in Computer Science from the Technical University of Catalonia (UPC). His areas of expertise include Linux, cluster management, parallel programming, storage, and hardware solutions, such as, iDataPlex, BladeCenter, and System x products.
Duncan Furniss is a Consulting IT Specialist for IBM in Canada. He currently provides technical sales support for PureFlex™, iDataPlex, BladeCenter, and System x products, and co-authored several IBM Redbooks publications. Duncan designed and provided oversight for the implementation of many large-scale solutions for HPC, distributed databases, and rendering of computer generated images. He is an IBM Certified IT Specialist and member of the IT Specialist Certification Review Board.
David Latino is a Senior HPC Solutions Architect for IBM Middle East & Africa. He has 10 years of experience in the HPC field. He led a wide spectrum of consulting projects, working with HPC users in academic research and industry sectors. His work covered many aspects of the HPC arena and he was technical leader for the design and implementation of multiple large HPC systems that appeared in the top500 list. He worked extensively on HPC application development, optimization, scaling, and performance benchmark evaluation, which resulted in several highly optimized application software packages. He also spent several years based at customer sites to train system administrators, users, and developers to manage and efficiently use IBM Blue Gene® systems.
Preface xi
Thanks to the following people for their contributions to this project:
From IBM Marketing:
򐂰 Mathieu Bordier 򐂰 Jill Caugherty 򐂰 Gaurav Chaudhry 򐂰 Kelly Chiu 򐂰 Jimmy Chou 򐂰 Chuck Fang 򐂰 Andrew Huang 򐂰 Camille Lee 򐂰 Brendan Paget 򐂰 Scott Tease 򐂰 Swarna Tsai 򐂰 Matt Ziegler
From IBM Development:
򐂰 David Brenchley 򐂰 Vincent Chao 򐂰 Kelly Chen 򐂰 Jason Cheng 򐂰 Marty Crippen 򐂰 Chris Hsieh 򐂰 Christina Hsu 򐂰 Jim Huang 򐂰 Cathy Lin 򐂰 Bruce Smith 򐂰 Brad Taylor 򐂰 Giant Tu 򐂰 Harold Wynkoop
From IBM Redbooks:
򐂰 Tam ikia B ar row 򐂰 Rich Conway
From IBMers around the world:
򐂰 Bill Champion 򐂰 Rick Koopman
A special thank you to Mathieu Bordier for hosting the team during our stay in Ta ip ei .
xii IBM NeXtScale System Planning and Implementation Guide
Now you can become a published author, too!
Here is an opportunity to spotlight your skills, grow your career, and become a published author—all at the same time! Join an ITSO residency project and help write a book in your area of expertise, while honing your experience using leading-edge technologies. Your efforts will help to increase product acceptance and customer satisfaction, as you expand your network of technical contacts and relationships. Residencies run from two to six weeks in length, and you can participate either in person or as a remote resident working from your home base.
Find out more about the residency program, browse the residency index, and apply online at this website:
http://www.ibm.com/redbooks/residencies.html
Comments welcome
Your comments are important to us!
We want our books to be as helpful as possible. Send us your comments about this book or other IBM Redbooks publications in one of the following ways:
򐂰 Use the online Contact us review Redbooks form found at this website:
http://www.ibm.com/redbooks
򐂰 Send your comments in an email to:
redbooks@us.ibm.com
򐂰 Mail your comments to:
IBM Corporation, International Technical Support Organization Dept. HYTD Mail Station P099 2455 South Road Poughkeepsie, NY 12601-5400
Preface xiii
Stay connected to IBM Redbooks
򐂰 Find us on Facebook:
http://www.facebook.com/IBMRedbooks
򐂰 Follow us on Twitter:
http://twitter.com/ibmredbooks
򐂰 Look for us on LinkedIn:
http://www.linkedin.com/groups?home=&gid=2130806
򐂰 Explore new Redbooks publications, residencies, and workshops with the
IBM Redbooks weekly newsletter:
https://www.redbooks.ibm.com/Redbooks.nsf/subscribe?OpenForm
򐂰 Stay current on recent Redbooks publications with RSS Feeds:
http://www.redbooks.ibm.com/rss.html
xiv IBM NeXtScale System Planning and Implementation Guide
© Copyright IBM Corp. 2013, 2014. All rights reserved. xv
Summary of changes
This section describes the technical changes made in this edition of the book and in previous editions. This edition might also include minor corrections and editorial changes that are not identified.
Summary of Changes for SG24-8152-01 for IBM NeXtScale System Planning and Implementation Guide as created or updated on July 7, 2014.
July 2014
This revision reflects the addition, deletion, or modification of new and changed information described below.
New information
򐂰 Added 1300W power supply efficiency values, Table 3-10 on page 40 򐂰 Added tables showing quantities of compute nodes supported based on
processor selection,, power supply selection, and input voltage, 3.3, “Supported compute nodes” on page 29
򐂰 Additional information on GPUs, 4.10, “GPU and coprocessor adapters” on
page 92
April 2014, Second Edition
This revision reflects the addition, deletion, or modification of new and changed information described below.
New information
򐂰 New PCIe Native Expansion Tray supporting GPUs and coprocessors 򐂰 New Intel Xeon Phi coprocessor and NVIDIA GPU adapter options 򐂰 New Intel Xeon E5-2600 v2 processor options 򐂰 New 1300W power supply option 򐂰 New models of the NeXtScale n1200 chassis with 1300W power supplies
xvi IBM NeXtScale System Planning and Implementation Guide
򐂰 New RDIMM memory options 򐂰 Support for 2.5-inch SSD options and other drive options 򐂰 Support for ServeRAID M5120 RAID controller for external SAS storage
expansion
© Copyright IBM Corp. 2013, 2014. All rights reserved. 1
Chapter 1. Introduction
IBM is introducing our next generation of scale out x86 servers, called NeXtScale System. In this chapter, we describe the client requirements that led us to its design, the computing environment it is meant to work in, and how this architecture was created to meet current and future business and technical challenges.
This chapter includes the following topics:
򐂰 1.1, “Evolution of data centers” on page 2 򐂰 1.2, “Executive Summary of IBM NeXtScale System” on page 3 򐂰 1.3, “Design points of the system” on page 6
1
2 IBM NeXtScale System Planning and Implementation Guide
1.1 Evolution of data centers
There is an increasing number of computational workloads that can be run on groups of servers; often referred to by such names as clusters, farms, or pools. This type of computing can be described as scale-out, although, as a convention, we refer to these groups as clusters. As the computing community’s proficiency with implementing and managing clusters improved, there is a trend to create large clusters, which are becoming known as hyper-scale environments.
In the past, when the number of servers in a computing environment was lower, to reduce application downtime, considerable hardware engineering effort and server cost was expended to create servers that were highly reliable. With clusters of servers, we strive to create a balance between the high availability technologies that are built in to every server, and reduce the cost and complexity of the servers, which allows more of them to be provisioned.
1.1.1 Density
As the number of servers in clusters grows and as data center real estate cost increases, the number of servers in a unit of space (also known as the compute density) becomes an increasingly important consideration. IBM NeXtScale System is designed to optimize density while addressing other objectives, such as, providing the best performing processors, minimizing the amount of energy that is used to cool the servers, and providing a broad range of configuration options.
1.1.2 Scale out applications
The following applications are among those that lend themselves to clusters of servers:
򐂰 High performance computing (HPC)
HPC is a general category of applications that are computationally complex, can deal with large data sets, or consist of vast numbers of programs that need to be run. Examples of computationally complex workloads include weather modeling or simulating chemical reactions. Comparing gene sequences is an example of a workload that involves large data sets. Image rendering for animated movies and Monte Carlo analysis for particle physics are examples of workloads where there are vast numbers of programs that need to be run. The use of several HPC clusters in a Grid architecture is an approach that gained popularity.
Chapter 1. Introduction 3
򐂰 Cloud services
Cloud services that are privately owned and those that are publicly available from managed service providers provide standardized computing resources from pools of homogeneous servers. If a consumer requires more or less server capacity, the servers are provisioned from or returned to the pools. This paradigm typically also includes consumer self service, and usage metering with some form of show back, charge back, or billing.
򐂰 Analytics
Distributed databases and extensive use of data mining, or analytics, is another use case that is increasing in prevalence and is being applied to a greater range of business and technical challenges.
1.2 Executive Summary of IBM NeXtScale System
NeXtScale System is IBM’s next generation dense, scalable computing system. The major components are the NeXtScale n1200 Enclosure and NeXtScale nx360 M4 compute node.
1.2.1 IBM NeXtScale n1200 Enclosure
The IBM NeXtScale System is based on a six-rack unit (6U) high chassis with 12 half-width bays, as shown in Figure 1-1.
Figure 1-1 Front of NeXtScale N1200 chassis, with 12 half-wide servers
4 IBM NeXtScale System Planning and Implementation Guide
Chassis power and cooling
The IBM NeXtScale n1200 Enclosure includes 10 hot swappable fans and six hot swappable power supplies. These are installed in the rear of the chassis, as shown in Figure 1-2.
Figure 1-2 Rear of IBM NeXtScale n1200 Enclosure
Six single-phase power supplies were chosen to enable power feeds from one or two sources of three-phase power.
Also, in the rear of the chassis is the Fan and Power Controller, which controls power and cooling aspects of the chassis.
1.2.2 IBM NeXtScale nx360 M4 compute node
The first server that is available for NeXtScale System is the nx360 M4 compute node. It fits in a half-width bay in the n1200 Enclosure, as shown in Figure 1-3 on page 5. On the front is the power button, status LEDs, and connectors. There is a full-height, half-length PCI Express card slot, and a PCI Express mezzanine card slot that uses the same mezzanine card type as our rack mount servers.
Chapter 1. Introduction 5
Figure 1-3 IBM NeXtScale nx360 M4
Inside, the nx360 M4 supports two Intel Xeon E5-2600 v2 series processors, eight DDR3 DIMMs, and a hard drive carrier. Hard disk drive carrier options include one 3.5-inch drive, two 2.5-inch drives, or four 1.8-inch solid-state drives. The server is shown in Figure 1-4.
Figure 1-4 IBM NeXtScale nx360 M4 with one 3.5-inch hard disk drive
6 IBM NeXtScale System Planning and Implementation Guide
1.3 Design points of the system
This section introduces some of the following design points that went into IBM NeXtScale System:
򐂰 The system is designed for flexibility.
The power supplies and fans, in the back of the chassis, are modular and hot swappable. The servers slide in and out of the front and have their cable connections at the front.
򐂰 The chassis is engineered to support devices that span multiple bays.
Compute node designs are not limited to a single 1U-high half-wide server. As with iDataPlex, servers can be augmented with trays that enable more features, such as, adapters or drives. Further systems that extend the design are in development.
򐂰 Fits in a standard rack
It is possible to install the IBM NeXtScale n1200 Enclosure into many standard 19-inch racks (which might require more cable routing brackets) and the rack can have a mixture of NeXtScale and other components. It is possible to start with a few servers and grow incrementally. Alternatively, you can have IBM install them into racks with switches and power distribution units and connect all the cables.
򐂰 IBM factory integration
Further configuration and testing are done when the systems are factory integrated. For more information about IBM factory integration, see Chapter 6, “Factory integration and testing” on page 153.
򐂰 The IBM NeXtScale System is focused on computational density.
Compared to an iDataPlex system with 84 servers in each iDataplex rack, with six NextScale chassis in each 42U standard rack (which leaves 6U per rack for more components), in the same floor tile configuration we can fit 28% more servers. With standard racks, clients should be able to design compact data center floor layouts for all their equipment.
򐂰 The system is also designed for simplicity.
Cable access to the server is from the front. The servers are directly accessed for their local console, management, and data networks, which eliminates contention.
򐂰 Another design objective was to use standard components.
The servers support standard PCI Express (Generation 3) adapters, and have RJ-45 copper Ethernet interfaces on board. The chipsets that are used in the server were selected with broad industry acceptance in mind.
Chapter 1. Introduction 7
1.4 This book
In this book, we compare NeXtScale to other systems, raising points to help you select the right systems for your applications. We then take an in-depth look at the chassis, servers, and fan and power controller (FPC). Next, we take a broader view, covering implementations at scale, reviewing racks and cooling.
We then describe IBM’s process for assembling and testing complete systems in to Intelligent Cluster™ solutions. Next, we provide information about managing the NeXtScale chassis and nodes. We finish by covering some of the software that is available from IBM that is commonly used in a solution with NeXtScale servers.
8 IBM NeXtScale System Planning and Implementation Guide
© Copyright IBM Corp. 2013, 2014. All rights reserved. 9
Chapter 2. Positioning
NeXtScale is ideal for fastest-growing workloads, such as, social media, analytics, technical computing, and cloud delivery, which are putting increased demands on data centers.
This chapter describes how IBM NeXtScale System is positioned in the marketplace compared with other systems that are equipped with Intel processors. The information helps you to understand the NeXtScale target audience and the types of workloads for which it is intended.
This chapter includes the following topics:
򐂰 2.1, “Market positioning” on page 10 򐂰 2.2, “IBM System x overview” on page 16 򐂰 2.3, “NeXtScale System versus iDataPlex” on page 17 򐂰 2.4, “NeXtScale System versus Flex System” on page 19 򐂰 2.5, “NeXtScale System versus rack-mounted servers” on page 20 򐂰 2.6, “Ordering and fulfillment” on page 21
2
10 IBM NeXtScale System Planning and Implementation Guide
2.1 Market positioning
The IBM NeXtScale System is a new x86 offering that introduces a new category of dense computing into the marketplace. IBM NeXtScale System includes the following key characteristics:
򐂰 Strategically, this is the next generation dense system from System x:
– A building block design that is based on a low function/low cost chassis. – Flexible compute node configurations that are based around a 1U
half-wide compute node supports various application workloads.
– A standard rack platform.
򐂰 Built for workloads that require density 򐂰 Not a replacement for, but complementary to, iDataPlex 򐂰 NeXtScale performs well in scale-out applications, such as, cloud, HPC, grid,
and analytics
򐂰 Is central in OpenStack initiatives for public clouds 򐂰 Available in standard System x configurator tools, including SSCT, Blue
Horizon, and x-config
IBM NeXtScale System includes the following key features: 򐂰 Supports up to seven chassis
1
in a 42U rack, which means up to a total of 84
systems and 2,016 processor cores in a standard 19-inch rack.
򐂰 High-value software stack for powerful scheduling, management, optimization
tools.
򐂰 Industry-standard components for flexibility, ease of maintenance, and
adoption.
򐂰 Approved for 40°C data centers, which lowers cooling costs. 򐂰 Available as single node, an empty or configured chassis, or in full racks. 򐂰 Can be configured as part of the IBM Intelligent Cluster processor for
complete pre-testing, configuration, and arrival ready to plug in.
򐂰 Compute nodes offer the fastest Intel Xeon processors (top-bin 130 W) with
new 1866 MHz memory.
򐂰 Supports 100 - 127 V and 200 - 240 V power. 򐂰 Standard form factor and components make it ideal for Business Partners.
1
Six chassis per rack are recommended because this leaves rack space for switches and cable routing. The use of seven chassis in a rack might require removal of the rack doors.
Chapter 2. Positioning 11
The customer who benefits the most from NeXtScale is an enterprise looking for a low-cost, high-performance computing system to start or optimize cloud, big data, Internet, and technical computing applications, which include the following uses:
򐂰 Large datacenters that require efficiency, density, scale, and scalability. 򐂰 Public, private, and hybrid cloud infrastructures. 򐂰 Data analytics applications, such as, customer relationship management,
operational optimization, risk and financial management, and enabling new business models.
򐂰 Internet media applications, such as, online gaming and video streaming. 򐂰 High-resolution imaging for applications ranging from medicine to oil and gas
exploration.
򐂰 “Departmental” uses in which a small solution can increase the speed of
outcome prediction, engineering analysis, and design and modeling.
2.1.1 Three key messages with NeXtScale
The three key messages about IBM NeXtScale System is that it is flexible, simple, and scalable, as shown in Figure 2-1.
Figure 2-1 IBM NeXtScale System key messages
Evenasmallclustercan changetheoutcome
Architectedfor performance,scale,and scalability
Maximumimpact/$
OptimizedSoftwareStack withPlatformComputing thepowerbeyondthe hardware
FLEXIBLE
Deliveredreadytorun
Channelandboxship capable
Optimizedforyourdata centertodayandreadyfor ittomorrow
BuiltonOpenStandards seamlessadoption
Onepartnumberunlocks IBMsserviceandsupport withIntelligentCluster
2
SIMPLE
1
SCALE
3
SingleArchitecture featuringNativeExpansion
Optimizedshared infrastructurewithout
compromisingperformance
Thebackisnowthefront
SimplySilver thepolishis itsessentialsonlydesign
12 IBM NeXtScale System Planning and Implementation Guide
IBM NeXtScale System is flexible in the following ways: 򐂰 Ordering and delivery
The question of how you want your NeXtScale System configuration to be ordered and delivered complex because there are many choices. Some clients want everything in parts so that they can mix and match to build what they want. Others want systems that they tested and approved to show that it is configured to their liking. Still others want complete solutions of racks to arrive ready to plug in. With NeXtScale, the choice is yours.
򐂰 A hardware design that allows for a mix of compute nodes.
The NeXtScale nx360 M4 compute node is a 1U half-wide server but is designed to be extended with the addition of various trays (what we call
Native Expansion) with which you can select the systems you need that is
based on the needs of the applications that you run.
򐂰 Fit it into your datacenter seamlessly in an IBM rack or most 19-inch standard
racks. The IBM NeXtScale n1200 Enclosure is designed to be installed in the IBM
42U 1100mm Enterprise V2 Dynamic Rack because it provides the best cabling features. However, the chassis can also be installed in many third-party, four-post, 19-inch racks. This ensures maximum flexibility when it comes to deploying NeXtScale System into your data center.
򐂰 IBM NeXtScale System is backed by leading IBM service and support no
matter how you buy it, where you use it, or what task you have it running.
򐂰 Support for open standards.
A client needs more than hardware to use IT. We designed NeXtScale to support an open stack of industry standard tools to allow clients that have existing protocols and tools to migrate easily to by using NeXtScale System.
The nx360 M4 compute node offers the Integrated Management Module II service processor and the n1200 Enclosure has the Fan and Power Controller. Both support the IPMI protocol for flexible and standards-based systems managements.
IBM NeXtScale System is simple in the following ways: 򐂰 A design that is based on the half-wide compute node.
The architecture of NeXtScale System revolves around a low-function chassis that hosts compute nodes. The design supports Native Expansion that allows seamless upgrades to add common functionality, such as, storage, graphics acceleration, or co-processing at the time of shipment or in the future.
Chapter 2. Positioning 13
򐂰 A chassis that includes shared fans and cooling.
The n1200 Enclosure supplies the cooling and power to maximize energy efficiency, but leaves management and connectivity to the compute nodes, which minimizes cost.
򐂰 Cables, connectors, and controls at the front.
With the exception of the power cords, all cabling is at the front of the chassis. All controls and indicators also are at the front. This configuration makes access, server swap-outs, and overall systems management easier.
Each compute node has a front connector for a local console for use if you need crash-cart access. Also, each compute node has a pull-out tab at the front for system and customer labeling needs, as shown in Figure 2-2.
Figure 2-2 Front of the IBM NeXtScale nx360 M4
Because the cables do not clog up the back of the rack, air flow is improved and thus energy efficiency also is improved. The harder the fans work to move air through server, the more power they use.
Your support staff who work in the data center can tell you the front of the rack is a much more enjoyable environment to spend time in because it might easily be 30°F (16°C) cooler at the front than at the back. People cannot stay in the rear of the rack that long before its no longer comfortable. Also, the front of rack is less noisy than the rear of the rack because of fan noise.
KVM port
Pull out label tab
14 IBM NeXtScale System Planning and Implementation Guide
Its also difficult to locate a single dense server in a row of dense racks and then go round the back to service the cabling. Having all of the cabling on the front simplifies and reduces the chances of mis-cabling or pulling the wrong server.
򐂰 Installation in a three-phase power data center.
The design of six power supplies per chassis allows seamless installation into data centers with three-phase power. With six supplies and two, three-phase feeds, we perfectly optimize and balance power delivery; no waste, no inefficiency.
򐂰 The compute nodes are unpainted metal.
Unlike every other x86 server IBM offers, these do not have a black front to them, which indicates simplicity and efficiency.
IBM NeXtScale System is scalable in the following ways: 򐂰 Scaling is for everyone
As we describe scale, do not assume its only for massive deployments; even a small one-chassis solution can change what users believe can be done.
No matter whether you start small and grow or start huge and grow enormously, NeXtScale is designed to be run and managed at scale as a single solution.
򐂰 NeXtScale System is built on what we learned about the financial aspects of
scale-out. Every decision that we made about the product was aimed at improving our
clients impact per dollar, whether that meant removing components that are not required or by selecting more energy efficient parts to reduce power usage and, therefore, power costs.
򐂰 The hardware can be used on its own or paired with a leadership stack that is
optimized for the work load. Whether you are running a high-performance computing (HPC) application
that is optimized with Platform Computing or run on top of an industry-standard, open-source tool, NeXtScale System makes it easy. Same for Cloud (OpenStack and other industry standard, open stacks, or IBM Smart Cloud); both are viable and bring value.
򐂰 Scalable to the container level.
IBM NeXtScale System can meet the needs of clients who want to add IT at the rack level or even at the container level. Racks can be fully configured, cabled, labeled, and programmed before they are shipped. IBM also can take configured racks and assemble them into complete, containerized solutions with power, cooling, and infrastructure delivered ready to go.
Chapter 2. Positioning 15
2.1.2 Optimized for workloads
IBM NeXtScale System is best-suited for the following primary workloads:
򐂰 Public and private cloud 򐂰 HPC and technical computing
Although these areas do share much in common, they have unique needs that are served with NeXtScale System, as shown in Figure 2-2.
Figure 2-3 IBM NeXtScale System: Optimized for cloud computing and HPC solutions
For cloud, the important factors are that we support the entire processor stack; so, no matter what our client goal is, we can support it with the right processor performance and cost point. The same is true of memory; we have cost and power-optimized choices and performance-optimized alternatives. The nodes contain the right onboard networking on board with 1 Gb NICs embedded, with options for up to four other high-speed fabric ports. With these features, the entire solution is designed to scale to any size.
HPC and technical computing have many of the same attributes as cloud; a key factor is the need for the top-bin 130 W processors. NeXtScale System can support top bin 130 W processors, which means more cores and higher frequencies than others.
Workload Fine Tuned Server Characteristics
Processor and Memory performance and choice Full
Intel stack support with memory for performance and/or cost optimization
Standard Rack optimized Fits into client data centers
seamlessly
Right sized IO Choice of networking options – 1Gb,
10Gb, or InfiniBand, all SDN ready
Infinitely Scalable from small to enormous grid
deployments all built on open standards
High energy efficiency means more impact/watt
Top bin Intel Xeon processors, large memory
bandwidth, and high IOPS for rapid transaction
processing and analytics
Workload optimized software stack with Platform
Computing and IBM xCAT
Architected for low latency with choice of high speed
fabric support
Supported as one part number no matter the size of
the solution and content with Intelligent Cluster
High Performance
Computing
Private Cloud Public Cloud
16 IBM NeXtScale System Planning and Implementation Guide
In addition to rock-solid hardware, we have a powerful software stack that is based on open standards and IBM added value offerings, such as, xCAT and Platform Computing. IBM provides a software stack to run on top of NeXtScale, including IBM General Parallel File System, GPFS™ Storage Server, xCAT, and Platform Computing, which provides scheduling, management, and optimization tools.
2.2 IBM System x overview
The world is evolving, and the way that our clients do business is evolving with it. That is why IBM has the broadest x86 portfolio in our history and is expanding even further to meet the needs of our clients, whatever those needs might be.
As shown in Figure 2-4 on page 17, the x86 server market is segmented on the following key product areas:
򐂰 High-end systems
IBM dominates this space with enterprise-class eX5 four-socket and up systems that offer unprecedented x86 performance, resiliency, and security.
򐂰 Blades and integrated systems
Integrated systems is a fast growing market where IBM adds value by packaging IBM software and hardware assets in a way that helps our clients optimize value from IBM systems.
򐂰 Dense systems
Dense systems, as with NeXtScale or iDataPlex, is a fast growing segment that is pushed by datacenter limitations and new workloads that require scale out architecture. These systems transformed how clients optimize space-constrained data centers with extreme performance and energy efficiency.
򐂰 High volume systems
The volume space is over half the total x86 server market and IBM has a broad portfolio of rack and tower servers to meet a wide range of client needs, from infrastructure to technical computing.
IBM solutions are aligned in a way to capture the value clients need from cloud, analytics, and technical computing. Supporting all of this are the IBM technology assets, such as, software and service and the deep integration that is necessary to support the entire portfolio.
Chapter 2. Positioning 17
Figure 2-4 System x Strategy delivers leadership products and solutions
2.3 NeXtScale System versus iDataPlex
Although iDataPlex and NeXtScale look different, many of the ideas and innovations we pioneered with iDataPlex remain in the new NeXtScale System.
When IBM introduced iDataPlex in 2008, we introduced a chassis that was dedicated to power and cool independent nodes. With NeXtScale system, we reuse the same principle, but we are extending it to bring more flexibility to the users.
The IBM NeXtScale n1200 Enclosure can now support up to 12 1U half-wide compute nodes, while the iDataPlex chassis can house only two 1U half-deep compute nodes. This allows IBM NeXtScale System to provide more flexibility to the user and mix in the chassis different types of nodes with different form factors.
The IBM NeXtScale n1200 Enclosure is designed to fit in the IBM 42U 1100mm Enterprise V2 Dynamic Rack, but it also fits in many standard 19-inch racks. Although iDataPlex can also be installed in a standard 19-inch rack, the use of iDataPlex racks and different floor layouts was required to make use of its high-density capability. IBM NeXtScale System brings more flexibility by allowing users to use standard 19-inch racks and does not require a special datacenter layout, which does not affect customer best practices and policies.
High-end systems
Broad portfolio to meet a wide range of client needs from infrastructure to technical computing
4 socket+ enterprise-class x86 performance, resiliency and security
Integration across IBM assets in systems and SW for maximum client optimization and value
Optimize space-constrained data centers with extreme performance and energy efficiency
IBM PureSystems
IBM eX5 Systems
IBM NeXtScaleIBM System x Rack & Tower
IBM Flex System
IBM BladeCenter
Blades/Integrated systems
Volume systems Dense systems
IBM iDataPlex
18 IBM NeXtScale System Planning and Implementation Guide
As with iDataPlex servers, NeXtScale servers support S3 mode. S3 allows systems to come back into full production from low-power state much quicker than a traditional power-on. In fact, cold boot normally takes about 270 second; with S3, it takes only 45 seconds. When you know that a system will not be used because of time of day or state of job flow, you can send it into a low-power state to save power and, when needed, bring it back online quickly.
In Table 2-1, the features of IBM NeXtScale System are compared to those features of IBM iDataPlex.
Table 2-1 Comparing IBM NeXtScale System to IBM iDataPlex
Feature iDataPlex NeXtScale Comments
Form factor Unique rack
1200 mm x 600 mm
Standard rack 600 mm x 1100 mm
NeXtScale System allows for lower-cost racks and, perhaps, racks customers already have.
Density in a standard 42U rack
Up to 42 servers Up to 84 servers (72 with
space for switches)
NeXtScale System can provide up to twice the server density when both types of servers are installed in a standard 42U rack.
Density in two consecutive floor tiles
a
a. Here we compare the density of servers that can be fitted in a single row of racks while using top-of-rack
switches. We use a single iDataPlex rack for iDataPlex servers that is 1200 mm wide, and we compare it with two standard racks for NeXtScale servers that also are 1200 mm wide.
84 servers 8 ToR switches (iDataPlex rack)
144 servers 12 ToR switches (two standard racks next to each other)
NeXtScale can provide up to 71% more servers per row when using top-of-rack (ToR) switches.
Density/400 sq. ft. (with Rear Door Heat Exchanger)
1,680 servers 84 servers/iDataPlex rack; four rows of five racks
2,160 servers 72 servers/standard rack; three rows of 10 racks
10x10 floor tiles 28% density increase because of standard rack layout.
Power/tile (front) 22 kW maximum
15 kW typical 42 servers + switches
37 kW maximum 25 kW typical 72 servers + switches
Similar power/server
GPU support Two GPUs per server in 2UTwo GPUs per server in 1U
effective space (GPU support planned)
GPU tray + base node = 1U. NeXtScale System has twice the density of iDataPlex.
Direct attached storage
None
b
b. None available with Intel E5-2600 v2 series processor.
Other storage-rich offerings that are planned include eight drives in 1U effective space
More flexibility with NeXtScale storage plan.
Direct water cooling Available Now NeXtScale System design
supports water cooling
Opportunity to optimize cost with NeXtScale.
Chapter 2. Positioning 19
2.4 NeXtScale System versus Flex System
Although NeXtScale System and Flex System are both “blade” architectures, they are different in their approach, as shown in Figure 2-5 and Table 2-2 on page 20. The key here is to understand the client philosophy. Features and approaches in Flex System and NeXtScale System appeal to different kinds of users.
Flex System has a wide ecosystem of compatible servers (x86 and POWER®), switches, and storage offerings. It is aimed at clients that are looking for an optimized and integrated solution that delivers better performance than competitive offerings.
NeXtScale System is an x86-only architecture and its aim is to be the perfect server for clients that require a scale-out infrastructure. NeXtScale System uses industry-standard components, including I/O cards and top-of-rack networking switches, for flexibility of choice and ease of adoption.
Figure 2-5 Comparing Flex System with NeXtScale System
Systems with integrated expertise to deliver lower management costs and a simplified user experience
Client needs: A pre-integrated, pre-
optimized, complete hardware and software platform with built-in stack management to develop and deliver new business applications quickly and cost effectively.
Designed for scale-out applications such as Technical Computing, Cloud, Grid and Analytics
Client needs: For clients looking to
build a customized solution that is delivered pre-integrated leveraging their existing management and networking stacks.
IBM PureFlex and Flex System
IBM NeXtScale System
20 IBM NeXtScale System Planning and Implementation Guide
Table 2-2 IBM Flex System and NeXtScale System Differentiations
2.5 NeXtScale System versus rack-mounted servers
Although the NeXtScale System compute nodes are included in a chassis, this chassis is only there to provide shared power and cooling.
As a consequence, the approach to design, buy, or upgrade a solution that is based on IBM NeXtScale System or regular 1U/2U rack-mounted servers is similar. In both cases, the full solution relies on separate networking components and can integrate seamlessly in an existing infrastructure.
The choice between IBM NeXtScale System and rack-mountable servers is first driven by application requirements because each form factor brings various advantages and limitations for particular workloads.
IBM Flex System differentiation IBM NeXtScale System differentiation
Integrated design offers flexibility (Flex System) and factory-integration (PureFlex System).
Flexibility of customization and integration of hardware.
10U chassis holding 14 nodes with shared power and cooling that is designed for multiple generations of technology.
6U chassis holding 12 servers with shared power and cooling designed for multiple generations of technology.
Heterogeneous node support: x86 and POWER.
x86 node support.
Full hardware redundancy; no single point of failure.
Designed for software redundancy; clustered approach.
Integrated switching, storage, and management in the chassis. All network connection are made via midplane. Support for 1 Gb/10 Gb Ethernet, FCoE, 8 Gb/16 Gb FC, and QDR/FDR InfiniBand. Custom form factor switches and adapters.
No Integrated switching in the chassis. Cables are routed to top of rack. Multiple brands of rack-based switches are supported: 1 GB/10 GB/40 Gb Ethernet, InfiniBand QDR/FDR, and 8 GB/16 GB Fibre Channel all installed external to the chassis. Standard PCI adapters and ToR switches.
Integrated shared storage optional with the Flex System V7000 Storage Node.
Integrated direct attached storage optional with the native expansion tray (planned).
Unified management via the Flex System Manager for all physical and virtual resources in chassis.
No unified management tools; management of node/ storage/switching handled via independent open standard, vendor-independent toolkits.
Chapter 2. Positioning 21
The 1U/2U rack-mounted servers and the IBM NeXtScale System feature the following main capabilities differences:
򐂰 Quantity of memory per node 򐂰 Quantity of PCIe slots per node 򐂰 Quantity of drives per node 򐂰 Support for high energy power consumption adapters, such as, GPU and
co-processors
For small installations or for servers requiring a large amount of memory, a large number of PCIe adapters the rack-mounted servers is the system of choice.
For medium to large installation of nodes that require up to 128 GB, IBM NeXtScale System should be the system of choice. It allows the users to reduce their initial cost of acquisition and their operating cost through higher density (up to 4X compared to 2U servers) and higher energy efficiency (because of the shared power and cooling infrastructure).
2.6 Ordering and fulfillment
We put careful attention on the NeXtScale System’s channel enablement. This allows not only IBM sellers to be talking about NeXtScale but also our valued partners. With iDataPlex, this was not something we saw; the product was not optimized for the partner’s use, sale, pricing, and so on. This is all different from NeXtScale because the product is set up the same as an HS23 or x3650, fully channel enabled, part of the Customer Transition Plan (CTP), xREF, and all the normal bid grids.
The IBM NeXtScale System can be fulfilled through regular channels or as a fully integrated solution with Intelligent cluster. For more information, see Chapter 6, “Factory integration and testing” on page 153.
22 IBM NeXtScale System Planning and Implementation Guide
IBM NeXtScale System is as easy to configure, order, and price as standard System x servers such as the x3650 M4 server, as shown in Table 2-3.
Table 2-3 Comparing configuration, ordering, and pricing tools
Tool System x iDataPlex NeXtScale
Presence in SSCT, Blue Horizon, x-config Ye s Ye s Ye s
Special bid in leads
Ye s Ye s Ye s
Available in IBM Intelligent Cluster
Ye s Ye s Ye s
Business Partners and distributors (Channel stocked)
Ye s N o Ye s
Grid level pricing
Ye s N o Ye s
Available to buy at ibm.com
Ye s N o Ye s
Reference Configurations - starter kits
Ye s N o Ye s
Platform-specific sales plays
Ye s N o Ye s
© Copyright IBM Corp. 2013, 2014. All rights reserved. 23
Chapter 3. IBM NeXtScale n1200
Enclosure
The foundation on which IBM NeXtScale System is built is the IBM NeXtScale n1200 Enclosure.
Designed to provide shared, high-efficiency power and cooling for up to 12 compute nodes, this chassis scales with your business needs. Adding compute, storage, or acceleration capability is as simple as adding nodes to the chassis. There is no built-in networking or switching capabilities, which requires no chassis-level management beyond power and cooling.
This chapter includes the following topics:
򐂰 3.1, “Overview” on page 24 򐂰 3.2, “Standard chassis models” on page 29 򐂰 3.3, “Supported compute nodes” on page 29 򐂰 3.4, “Power supplies” on page 38 򐂰 3.5, “Fan modules” on page 42 򐂰 3.6, “Midplane” on page 45 򐂰 3.7, “Fan and Power Controller” on page 47 򐂰 3.9, “Specifications” on page 57
3
24 IBM NeXtScale System Planning and Implementation Guide
3.1 Overview
The IBM NeXtScale n1200 Enclosure is a 6U next-generation dense server platform with integrated Fan and Power Controller. The n1200 is designed to efficiently power and cool up to 12 1U half-wide compute nodes, with which clients can install in a standard 42U 19-inch rack that is twice the number of servers per rack-U space that is compared to traditional 1U rack servers.
Figure 3-1 IBM NeXtScale n1200 Enclosure with 12 compute nodes
The founding principle behind IBM NeXtScale System is to allow clients to adopt this new hardware with minimal or no changes to their existing data center infrastructure, management tools, protocols, and best practices.
The enclosure looks similar to an IBM BladeCenter or IBM Flex System chassis but it is different as there is no consolidated management interface or integrated switching. An exploded view of the components of the chassis is shown in Figure 3-2 on page 25.
Chapter 3. IBM NeXtScale n1200 Enclosure 25
Figure 3-2 IBM NeXtScale n1200 Enclosure components
The IBM NeXtScale n1200 Enclosure includes the following components:
򐂰 Up to 12 compute nodes 򐂰 Six power supplies each separately powered 򐂰 A total of 10 fan modules in two cooling zones 򐂰 One Fan and Power Controller
Shipping
bracket kit
Midplane
Top cover
Fan modules
Fan and Power Controller
Power supplies
Rail kit
Node filler
Lift handles
26 IBM NeXtScale System Planning and Implementation Guide
3.1.1 Front components
The IBM NeXtScale n1200 Enclosure supports up to 12 1U half-wide compute nodes, as shown in Figure 3-3.
All compute nodes are front accessible with front cabling as shown in Figure 3-3. From this angle, the chassis looks to be simple because it was designed to be simple, low cost, and efficient.
Figure 3-3 IBM NeXtScale n1200 Enclosure Front View with 12 compute nodes
This new enclosure not only supports dense, high-performance compute nodes, but also expanded compute nodes with more I/O slots for adapters, GPUs, and coprocessors or other drive bays.
By using this capability, clients can access some powerful IT inside a simple and cost effective base compute node that can be expanded to create rich and dense storage or acceleration solutions without the need for any exotic components, midplanes, or high-cost connectors.
Bay 1
Bay 3
Bay 5
Bay 7
Bay 9
Bay 11
Bay 2
Bay 4
Bay 6
Bay 8
Bay 10
Bay 12
Chapter 3. IBM NeXtScale n1200 Enclosure 27
3.1.2 Rear components
The n1200 provides shared high-efficiency power supplies and fan modules. As with IBM BladeCenter and IBM Flex System, the NeXtScale System compute nodes connect to a midplane, but this connection is for power and control only; the midplane does not provide any I/O connectivity.
Figure 3-4 shows the major components that are accessible from the rear of the chassis.
Figure 3-4 IBM NeXtScale n1200 Enclosure rear view
At the rear of the chassis, the following types of components are accessible: 򐂰 Power supplies
The IBM NeXtScale n1200 Enclosure has a six-power supply design, all in one power domain. This configuration allows clients with 100 V - 240 V utility power (in single or three-phase) to power up the chassis by using the infrastructure they already have. For three-phase power, the phases are split in the PDU for single phase input to the chassis power supplies.
For more information about the power supplies, see 3.4, “Power supplies” on page 38.
6x Power supplies Fan and Power Controller 10x Fan Modules
28 IBM NeXtScale System Planning and Implementation Guide
򐂰 Fan modules
Also shared in the chassis are 10 80 mm fan modules, five in each of the two cooling zones.
The fan modules and PSUs in the chassis provide shared power and cooling for all the installed nodes by using fewer components and with less power than traditional systems.
For more information about the fan modules, see 3.5, “Fan modules” on page 42.
򐂰 Fan and Power Controller
The Fan and Power Controller (FPC) module is the management device for the chassis and, as its name implies, controls the power and cooling features of the enclosure.
For more information about the FPC, see 3.7, “Fan and Power Controller” on page 47.
You might notice that the enclosure does not contain space for network switches. All I/O is routed directly out of the servers to top-of-rack switches. This configuration provides choice and flexibility and keeps the IBM NeXtScale n1200 Enclosure flexible and low cost.
3.1.3 Fault tolerance features
The chassis implements a fault tolerant design. The following components in the chassis enable continued operation if one of the components fails:
򐂰 Power supplies
The power supplies support a single power domain that provides DC power to all of the chassis components. If a power supply fails, the other power supplies can continue to provide power.
򐂰 Fan modules
The fan modules provide cooling to all of the chassis components. The power supplies have their own fans to provide the cooling. Each fan module has a dual rotor (blade) dual motor fan. One of the motors within the fan module can fail and the remaining continue to operate. If a fan fails, the chassis continues operating and the remaining fans increase in speed to compensate.
Power policies: The power management policy that you implemented for the chassis determines the affect on chassis operation if there is a power supply failure. Power policies can be N+N, N+1, or no redundancy. Power policies are managed by the FPC.
Chapter 3. IBM NeXtScale n1200 Enclosure 29
򐂰 FPC
The FPC enables the Integrated Management Module to monitor the fans and control fan speed. If the FPC fails, the enclosure fans ramp up to maximum, but all systems continue to operate by using the existing power management policy.
3.2 Standard chassis models
The standard chassis model is listed in Table 3-1. The chassis is also available via the configure-to-order (CTO) processor.
Table 3-1 Standard enclosure models
The IBM NeXtScale n1200 Enclosure ships with the following items:
򐂰 Rail kit 򐂰 One Console breakout cable, also known as a KVM dongle (00Y8366) 򐂰 A Torx-8 (T8) screwdriver for use with components, such as, the drive cage,
which is mounted on the rear of the chassis
򐂰 One AC power cord for each power supply installed, 1.5m 10A, IEC320 C14
to C13 (part number 39Y7937)
3.3 Supported compute nodes
The IBM NeXtScale nx360 M4 is the only compute node that is supported in the n1200 enclosure. However, the number of compute nodes that can be powered on depends on the following factors:
򐂰 The power policy that is selected (N+N, N+1, or no redundancy) 򐂰 The AC input voltage 򐂰 The components that are installed in each compute node (such as, processor,
memory, drives, and PCIe adapters)
Model Description Fan Modules
(standard/max)
Power Supplies (standard/max)
5456-A2x IBM NeXtScale n1200 Enclosure 10x 80mm / 10 6x 900 W / 6
5456-A3x IBM NeXtScale n1200 Enclosure 10x 80mm / 10 2x 1300 W / 6
5456-A4x IBM NeXtScale n1200 Enclosure 10x 80mm / 10 6x 1300 W / 6
30 IBM NeXtScale System Planning and Implementation Guide
򐂰 UEFI settings
To size for a specific configuration, you can use the IBM Power Configurator that is available at this website:
http://ibm.com/systems/bladecenter/resources/powerconfig.html
The following show the number of nodes that can be operated with no performance compromise within the chassis depending on the power policy required. The tables are as follows:
򐂰 1300 W power supplies:
– 200-240V AC input, no GPU Trays: Table 3-2 – 200-240V AC input, GPU Trays with 130 W GPUs: Table 3-3 on page 32 – 200-240V AC input, GPU Trays with 225 W GPUs: Table 3-4 on page 33 – 200-240V AC input, GPU Trays with 235 W GPUs: Table 3-5 on page 34 – 200-240V AC input, GPU Trays with 300 W GPUs: Table 3-6 on page 35
򐂰 900 W power supplies:
– 200-240V AC input, no GPU Trays: Table 3-7 on page 36 – 100-127V AC input, no GPU Trays: Table 3-8 on page 37
Table 3-2 shows the quantity supported of compute nodes with six 1300 W power supplies installed in the chassis.
Table 3-2 Number of compute nodes supported (high-line AC Input, with 6 x 1300 W PSUs)
Power Configurator: Use the IBM Power Configurator to determine an accurate power model for your configuration.
http://ibm.com/systems/bladecenter/resources/powerconfig.html
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
12 12 12 12
2
12 12 12 12
60 W 1
12 12 12 12
2
12 12 12 12
70 W 1
12 12 12 12
2
12 12 12 12
Chapter 3. IBM NeXtScale n1200 Enclosure 31
Table 3-3 shows the quantity supported of compute nodes with GPU trays attached and two 130 W GPUs installed. Only 1300 W power supplies are supported and only 200-240V AC input supported. The supported 130 W GPU is:
򐂰 NVIDIA GRID K1, 00J6160
80 W 1 12 12 12 12
2
12 12 12 12
95 W 1
12 12 12 12
2
12 12 10 12
115 W 1
12 12 12 12
2
12 12 8 12
130 W 1
12 12 12 12
2
12 12 7 11
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
N+1 N+N
N+N with OVS
a
Note: Some cells indicate two numbers (for example “5 + 1”). This indicates support for a mixture of servers with and without the GPU Tray:
򐂰 First number: Number of servers with a GPU Tray attached and two GPUs
installed
򐂰 Second number: Number of servers without a GPU Tray attached.
For example, “5 + 1” means Supported combination is
5 servers with the GPU
Tray attached, plus
1 server without a GPU Tray attached. In such a
configuration, the 1 remaining server bay in the chassis must remain empty.
32 IBM NeXtScale System Planning and Implementation Guide
Table 3-3 Number of compute nodes supported each with two 130 W GPUs installed (6 x 1300 W PSUs)
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
6 6 6 6
2
6 6 6 6
60 W 1
6 6 6 6
2
6 6 6 6
70 W 1
6 6 6 6
2
6 6 6 6
80 W 1
6 6 6 6
2
6 6 6 6
95 W 1
6 6 6 6
2
6 6 5 + 1
b
b. See shaded box on page 31
6
115 W 1
6 6 6 6
2
6 6 5 6
130 W 1
6 6 5 + 1
b
6
2
6 6 4 + 1
b
5 + 1
b
Chapter 3. IBM NeXtScale n1200 Enclosure 33
Table 3-4 on page 33 shows the quantity supported of compute nodes with GPU trays attached and two 225 W GPUs installed. Only 1300 W power supplies are supported and only 200-240V AC input supported. The supported 225 W GPUs are:
򐂰 Intel Xeon Phi 5110P, 00J6163 򐂰 NVIDIA Tesla K10, 00D4192 򐂰 NVIDIA GRID K2, 00J6161 򐂰 NVIDIA Tesla K20X 00J6165
Table 3-4 Number of compute nodes supported each with two 225 W GPUs installed (6 x 1300 W PSUs)
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
6 6 5 + 1
b
b. See shaded box on page 31
6
2
6 6 5 6
60 W 1
6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
70 W 1 6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
80 W 1 6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
95 W 1 6 6 4 + 2
b
6
2
6 6 4 5
115 W 1
6 6 4 + 1
b
5 + 1
b
2 6 6 3 + 1
b
4 + 1
b
130 W 1 6 6 4 + 1
b
5
2
6 6 3 + 1
b
4 + 1
b
34 IBM NeXtScale System Planning and Implementation Guide
Table 3-5 on page 34 shows the quantity supported of compute nodes with GPU trays attached and two 235 W GPUs installed. Only 1300 W power supplies are supported and only 200-240V AC input supported. The supported 235 W GPU is:
򐂰 NVIDIA Tesla K40, 00FL133
Table 3-5 Number of compute nodes supported each with two 235 W GPUs installed (6 x 1300 W PSUs)
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
6 6 5 + 1
b
b. See shaded box on page 31
6
2
6 6 4 + 1
b
6
60 W 1
6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
70 W 1 6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
80 W 1 6 6 5 6
2
6 6 4 + 1
b
5 + 1
b
95 W 1 6 6 4 + 2
b
5 + 1
b
2 6 6 4 5
115 W 1
6 6 4 + 1
b
5 + 1
b
2 6 6 3 + 1
b
4 + 1
b
130 W 1 6 6 4 5
2
6 6 3 + 1
b
4
Chapter 3. IBM NeXtScale n1200 Enclosure 35
Table 3-5 on page 34 shows the quantity supported of compute nodes with GPU trays attached and two 300 W GPUs installed. Only 1300 W power supplies are supported and only 200-240V AC input supported. The supported 300 W GPU is:
򐂰 Intel Xeon Phi 7120P, 00J6162
Table 3-6 Number of compute nodes supported each with two 300 W GPUs installed (6 x 1300 W PSUs)
Compute nodes Power policy - 6 x 1300W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
6 6 4 + 2
b
b. See shaded box on page 31
5 + 1
b
2 6 6 4 5
60 W 1
6 6 4 5
2
6 6 3 + 2
b
4 + 2
b
70 W 1 6 6 4 5
2
6 6 3 + 2
b
4 + 2
b
80 W 1 6 6 4 5
2
6 6 3 + 2
b
4 + 2
b
95 W 1 6 6 4 4 + 2
b
2 6 6 3 + 1
b
4 + 1
b
115 W 1 6 6 3 + 2
b
4 + 2
b
2 6 5 + 1
b
3 3 + 2
b
130 W 1 6 6 3 + 2
b
4 + 1
b
2 6 5 + 1
b
3 3 + 2
b
36 IBM NeXtScale System Planning and Implementation Guide
Table 3-7 shows the quantity supported of compute nodes with six 900 W power supplies installed in the chassis with those power supplies connected to a 200-240V (high-line) AC input.
Table 3-7 Number of compute nodes supported (high-line AC Input, with 6 x 900 W PSUs)
Compute nodes Power policy - 6 x 900W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
High line AC input (200-240 V)
50 W 1
12 12 12 12
2
12 12 11 12
60 W 1
12 12 12 12
2
12 12 10 12
70 W 1
12 12 12 12
2
12 12 8 11
80 W 1
12 12 11 12
2
12 12 8 9
95 W 1
12 12 10 12
2
12 12 6 10
115 W 1
12 12 8 10
2
12 10 5 8
130 W 1
12 12 7 9
2
10 8 4 7
Chapter 3. IBM NeXtScale n1200 Enclosure 37
Table 3-8 shows the quantity supported of compute nodes with six 900 W power supplies installed in the chassis with those power supplies connected to a 100-127 V (low-line) AC input.
Table 3-8 Number of compute nodes supported (low-line AC Input, with 6 x 900 W PSUs)
Compute nodes Power policy - 6 x 900W power supplies
CPU TDP
Number of CPUs
Non-redundant or N+1 with OVS
a
a. OVS (oversubscription) of the power system allows for more efficient use of the available system
power.
N+1 N+N
N+N with OVS
a
Low line AC input (100-127 V)
50 W 1
12 12 9 11
2
12 12 6 10
60 W 1
12 12 7 9
2
12 9 5 7
70 W 1
12 12 7 9
2
12 9 5 7
80 W 1
12 12 6 8
2
10 9 5 7
95 W 1
12 11 6 7
2
9 7 4 6
115 W 1
11 9 5 6
2
7 6 3 5
130 W 1
9 8 4 5
2
6 5 3 4
38 IBM NeXtScale System Planning and Implementation Guide
3.4 Power supplies
The IBM NeXtScale n1200 Enclosure supports up to six high-efficiency autoranging power supplies. The standard model includes all six power supplies. A single power supply is shown in Figure 3-5.
Figure 3-5 900 W power supply
Table 3-9 lists the ordering information for the supported power supplies.
Table 3-9 Power supplies
The power supply options have the following features: 򐂰 Supports N+N or N+1 Power Redundancy, or Non-redundant power
configurations to support higher density
򐂰 Power management controller and configured through the Fan and Power
Controller
򐂰 Integrated 2500 RPM fan 򐂰 80 PLUS Platinum certified 򐂰 Built-in overload and surge protection
Part number
Feature code
Description Min / Max
supported
Chassis model where used
00Y8569 A41T CFF 900 W Power Supply 6 / 6 A2x
00Y8652 A4MM CFF 1300W Power Supply 2 / 6 A3x, A4x
Chapter 3. IBM NeXtScale n1200 Enclosure 39
900W power supply specifications:
򐂰 Supports dual-range voltage: 100 - 240 V 򐂰 100 - 127 (nominal) V AC; 50 or 60 Hz; 6.8 A (maximum) 򐂰 200 - 240 (nominal) V AC; 50 or 60 Hz; 5.0 A (maximum)
1300W power supply specifications:
򐂰 Supports high-range voltage only: 200 - 240 V 򐂰 200 - 240 (nominal) V AC; 50 or 60 Hz; 6.9 A (maximum)
The location and numbering of the power supplies are shown in Figure 3-6.
Figure 3-6 IBM NeXtScale n1200 Enclosure rear view with power supply numbering
The power supplies that are used in IBM NeXtScale System are hot-swap, high-efficiency 80 PLUS Platinum power supplies that are operating at 94% efficiency. The efficiency varies by load, as shown in Table 3-10. The 80 PLUS report is available at this website:
http://www.plugloadsolutions.com/psu_reports/IBM_7001700-XXXX_900W_SO-5 71_Report.pdf
1300W power supply: The 1300W power supply does not support low-range voltages (100-127V).
Fan and Power
Controller
Static IP User Name Password
192.168.0.100 USERID PASSW0RD
Zero
Default Information:
Event Log-AlternateAccess
via USB Key
USB key
Fan Baysand Power Supply
6 5 8
7
10
9
2
4
46
3
3
5
1
21
http://ibm.co/19pHxV8
5 3
1
6 4
2
40 IBM NeXtScale System Planning and Implementation Guide
Table 3-10 Power efficiencies at different load levels
The 80 PLUS performance specification is for power supplies that are used within servers and computers. To meet the 80 PLUS standard, the power supply must have an efficiency of 80% or greater, at 20%, 50%, and 100% of rated load with PF of 0.9 or greater. The standard has several grades, such as, Bronze, Silver, Gold, and Platinum. For more information about the 80 PLUS standard, see this website:
http://www.80PLUS.org
The power supplies receive electrical power from a 100 V - 127 V AC or 200 v ­240 V AC power source and convert the AC input into DC outputs. The power supplies can autorange within the input voltage range.
There is one common power domain for the chassis that distributes DC power to each of the nodes and modules through the system midplane.
DC redundancy is achieved when there is one more power supply available than is needed to provide full power to all chassis components. AC redundancy is achieved by distributing the AC power cord connections between independent AC circuits. For more information, see 5.1, “Power planning” on page 116.
Each power supply includes presence circuitry, which is powered by the midplane. This allows the FPC to recognize when power supplies are installed in the enclosure but are not powered by the AC supply.
The power supplies support oversubscription. By using oversubscription, users can make the most of the extra power from the redundant power supplies when the power supplies are in healthy condition. This means you can use the power capacity of all installed power supplies while still preserving power supply redundancy if there is a power supply failure. For more information, see “Power supply oversubscription” on page 54.
20% load 50% load 100% load
80 PLUS Platinum standard 90.00% 94.00% 91.00%
NeXtScale n1200 900 W power supply 92.96% 94.15% 92.53%
NeXtScale n1200 1300 W power supply 93.97% 94.54% 92.65%
Use with 110-127 V AC: When low input voltage (100 V - 127 V AC) is used, the power supply is limited to 600 W.
Chapter 3. IBM NeXtScale n1200 Enclosure 41
As shown in Figure 3-5 on page 38, the following LEDs are on each power supply:
򐂰 AC power LED
When this LED is lit (green), it indicates that AC power is supplied to the power supply.
򐂰 DC power LED
When this LED is lit (green), it indicates that DC power is supplied from the power supply to the chassis midplane.
򐂰 Fault LED
When this LED is lit (yellow), it indicates that there is a fault with the power supply.
Removing a power supply: To maintain proper system cooling, do not operate the IBM NeXtScale n1200 Enclosure without a power supply (or power supply filler) in every power supply bay. Install a power supply within 1 minute of the removal of a power supply.
42 IBM NeXtScale System Planning and Implementation Guide
3.5 Fan modules
The IBM NeXtScale n1200 Enclosure supports 10, 80 mm fan modules. All fans modules are at the rear of the chassis and are numbered as shown in Figure 3-7.
Figure 3-7 IBM NeXtScale n1200 Enclosure rear view with fan bay numbering
The fan modules provide cooling to the compute nodes. The fan modules have a dual-rotor design for high efficiency and high reliability. Air flow is front-to-back.
Ordering information for the fan modules is shown in Table 3-11.
Table 3-11 Fan Modules
Part number
Feature code
Description Maximum
supported
Chassis model where used
00Y8570 A41F n1200 Fan Module 10 A2x
Fan and Power
Controller
Static IP User Name Password
192.168.0.100 USERID PASSW0RD
Zero
Default Information:
Event Log-AlternateAccess
via USB Key
USB key
Fan Baysand Power Supply
6 5 8
7
10
9
2
4
46
3
3
5
1
21
http://ibm.co/19pHxV8
7
5
2134
6
8910
Chapter 3. IBM NeXtScale n1200 Enclosure 43
The 80 mm fan module is shown in Figure 3-8.
Figure 3-8 80 mm fan module
Fan module controls and indicators
Each fan module has a Fault LED. When this LED is lit (yellow), it indicates that the fan module failed.
The fan modules are not dedicated to cool a specific node. If there is a fan module failure, the remaining functional fans speed up (if required) under the control of the FPC to provide sufficient cooling to the chassis elements.
There are two logical cooling zones in the enclosure, as shown in Figure 3-9 on page 44. Five fan modules on each side correspond to the six compute nodes on that same side of the chassis.
44 IBM NeXtScale System Planning and Implementation Guide
Figure 3-9 Cooling zones on the chassis
For each of the cooling zones (zone 1 or zone 2), the FPC sets the respective fans for the corresponding nodes to the appropriate cooling values that are required to cool those nodes. The FPC varies the speeds of the fans in zone 1 and zone 2 by at most a 20% difference to avoid unbalanced air flow distribution.
Fan removal: To maintain proper system cooling, do not operate the IBM NeXtScale n1200 Enclosure without a fan module (or fan module filler) in every fan module bay. Install a fan module within 1 minute of the removal of a fan module.
Cooling zone 2
Cooling zone 1
Chapter 3. IBM NeXtScale n1200 Enclosure 45
3.6 Midplane
The enclosure midplane is the bridge to connect the compute nodes with the power supplies, fan modules, and the FPC. Figure 3-10 shows the front and rear of the midplane.
Figure 3-10 Front and rear view of the n1200 Enclosure Midplane Assembly
The midplane is used to provide power to all elements in the chassis. It also provides signals to control fan speed, power consumption, and node throttling.
The midplane was designed with no active components to improve reliability and minimize serviceability. Unlike BladeCenter, for example, the midplane is removed by removing a cover on the top of the chassis.
Front view Rear view
46 IBM NeXtScale System Planning and Implementation Guide
Figure 3-11 shows the connectivity of the chassis components through the midplane.
Figure 3-11 Midplane connectivity
Midplane
Fan and Power Controller
Servers
Power Supplies
Fan
Modules
Power Supplies
Fan
Modules
I2C
Mux
12
FPC
12
ARM SOC
RJ45
I2C
Mux
USB
Chapter 3. IBM NeXtScale n1200 Enclosure 47
3.7 Fan and Power Controller
FPC controls the power budget, gives the power permission to each node, and controls the speed of the fans. The FPC is installed inside the chassis and is accessible from the rear of the chassis, as shown in Figure 3-12. The FPC is a hot-swap component, as indicated by the orange handle.
Figure 3-12 Rear view of the chassis that shows the location of the FPC
48 IBM NeXtScale System Planning and Implementation Guide
3.7.1 Ports and connectors
The FPC provides integrated systems management functions. The user interfaces (browser and CLI) are accessible remotely via the 10/100 Mbps Ethernet port.
Figure 3-13 shows the FPC and its LEDs.
Figure 3-13 FPC
The FPC has the following LEDs and connector that you can use to obtain status information and restart the FPC:
򐂰 Power LED
When this LED is lit (green), it indicates that the FPC has power.
򐂰 Heartbeat LED
When this LED is lit (green), it indicates that the FPC is actively controlling the chassis.
򐂰 Locator LED
When this LED is lit (blue), it indicates the chassis location in a rack.
򐂰 Check log LED
When this LED is lit (yellow), it indicates that a system error occurred.
򐂰 Ethernet port activity LED
When this LED is flashing (green), it indicates that there is activity through the remote management and console (Ethernet) port over the management network.
Power LED
Heartbeat LED
Locator LED
USB key
Ethernet port link LED
Ethernet port activity LED
Check log LED
Chapter 3. IBM NeXtScale n1200 Enclosure 49
򐂰 Ethernet port link LED
When this LED is lit (green), it indicates that there is an active connection through the remote management and console (Ethernet) port to the management network.
򐂰 Remote management and console (Ethernet) connector
The remote management and console RJ45 connector is the management network connector for all chassis components. This 10/100 Mbps Ethernet connector is connected to the management network through a top-of-rack switch.
3.7.2 Internal USB memory key
The FPC also includes a USB key that is housed inside the unit, as shown in Figure 3-14.
Figure 3-14 Internal view of the FPC
Note: The FPC is not a point of failure for the chassis by design. If the FPC fails, the compute nodes, power supplies, and fans remain functional to keep the systems running. The power capping policies that are set to the chassis and compute nodes remain in place. The fan speeds up to maximum speed to provide maximum cooling protection for the compute nodes until the FPC is replaced.
50 IBM NeXtScale System Planning and Implementation Guide
The USB key saves the following information:
򐂰 Event log 򐂰 Enclosure configuration data:
– PSU redundancy setting – Oversubscription mode setting – Chassis/node-level power capping value and settings – Power restore policy – Acoustic mode setting
򐂰 Midplane vital product data (VPD)
If the FPC fails and must be replaced, users can restore the configuration data to the new FPC by transferring the USB from the old unit to the new unit.
3.7.3 Overview of functions
The FPC performs the following functions: 򐂰 Controls the power and power management features of the chassis, compute
nodes, and power supplies. The FPC prevents a compute node from powering on if there is insufficient power.
򐂰 Controls the cooling of the chassis. The FPC ramps up fan speeds if
conditions require more cooling (or slow down the fans to conserve energy if higher fan speeds are not required).
򐂰 Provides the following user interfaces:
– Web interface – IPMI command line (for external management tools, such as, xCAT)
򐂰 Allows you to select one of the following power supply redundancy policies:
– N+1, where one power supply is redundant and allows for a single power
supply to fail without any loss of function.
– N+N, where half the power supplies are redundant backups of the other
half. This is useful if you have two power utility sources and want the chassis to survive the failure of one of the utility sources.
– Non-redundant, which maximizes the power that is available to the
compute nodes at the expense of power supply redundancy.
– Oversubscription, which can be enabled with N+1 and N+N policies.
򐂰 Support the updating of FPC firmware. 򐂰 Monitors and reports fan, power, and chassis status and other failures with
event log and corresponding LEDs.
Chapter 3. IBM NeXtScale n1200 Enclosure 51
3.7.4 Web GUI interface
Through the FPC web interface, the user or system administrator can perform the following tasks. For more information about the FPC web interface, see 7.2.1, “FPC web browser interface” on page 180:
򐂰 View summary of elements status:
– Front and rear view of the chassis – Compute nodes location and status – FPC, power supplies, fan modules location, and status
򐂰 View current power usage:
– Voltage overview of the chassis – Total chassis power consumption (AC-in) – Total PSU power consumption (DC-out) – Total fans power consumption – Per-node power consumption – Power supply fan speeds
򐂰 View and set power supply redundancy and oversubscription:
– Select N, N+1, or N+N mode – Enable or disable oversubscription mode
򐂰 View and set power capping:
– Node level: Set value within a defined range for each node separately, or
choose between one of the three predefined modes.
– Chassis level: Set value within a defined range for the enclosure, or
choose between one of the three predefined modes.
򐂰 View and set power restore policy: Enable or disable (for each node or
chassis)
򐂰 View current fan speeds 򐂰 View and set Acoustic mode (three modes) 򐂰 View Chassis, Midplane, and FPC vital product data (VPD) details 򐂰 View, save, and clear the system event log 򐂰 View and set network configuration:
– SMTP configuration – SNMP traps and email alert configuration – Host name, DNS, Domain, IP, and IP version configuration – SNMP traps email alert configuration – Web server (http or https) ports configuration
򐂰 Perform a Virtual Reset or Virtual Reseat of each compute node
52 IBM NeXtScale System Planning and Implementation Guide
򐂰 Set Locator (Identify) LED to on, off, or blink 򐂰 Turn off Check Log LED 򐂰 Back up and restore FPC configuration to USB key; reset to default 򐂰 Perform firmware update 򐂰 Set date and time 򐂰 Perform user account management
3.8 Power management
The FPC controls the power on the IBM NeXtScale n1200 Enclosure. If there is sufficient power available, the FPC allows a compute node to be powered on.
The power permission includes the following two-step process:
1. Pre-boot (BMC stage) inventory power (standby power) is pre-determined based on the node type.
2. Post-boot (UEFI stage) inventory power is a more accurate estimation of the node’s maximum power usage that is based on power maximizer test. The following values are generated:
– Maximum power usage value under stressed condition. – Maximum power usage value under stressed condition when P-state is
capped at the lowest level.
The FPC uses these values to compare the total node power and total available power to determine power-on and boot permissions.
3.8.1 Power Restore policy
By using the Power Restore policy, you can specify whether you want the compute nodes to restart when a chassis AC power is removed and restored. This is similar to the Automatic Server Restart (ASR) feature of many System x servers. For more information about the Power Restore Policy, see “Power Restore Policy tab” on page 189.
When Power Restore policy is enabled, if the compute node was powered on before the AC power cycle, the FPC turns the compute node back on automatically when power is restored.
Chapter 3. IBM NeXtScale n1200 Enclosure 53
However, a compute node is restarted only if the FPC determines there is sufficient power available to power on the server, which is based on the following factors:
򐂰 Number of working power supplies 򐂰 Power policy that is enabled 򐂰 The oversubscription policy
If there is insufficient power, some compute nodes are not powered on. The nodes are powered on based on their power usage, lowest first. The objective is to maximize the number of nodes that can be powered on in a chassis. The power-on sequence nominally takes about 2 minutes with most of that time spent by the nodes running a power maximizer. After that is complete, the FPC can quickly release the nodes to continue to boot.
3.8.2 Power capping
User can choose chassis-level capping or saving or node-level capping or saving through the power cap configuration options. Power capping allows users to set a wattage limit on power usage. When it is applied to an individual node, the node power consumption is capped at an assigned level. When it is applied to a chassis, the whole chassis power usage is capped. When power saving is enabled, an individual node or all nodes (chassis level) run in modes of different throttling level, depending on the modes that are chosen.
3.8.3 Power supply redundancy modes
The FPC offers the following modes from which to choose: 򐂰 No-redundancy mode
The loss of any power supply can affect the system’s operation or performance. If the chassis is evaluated to be vulnerable, because of the failure of one or multiple power supplies, throttle signals are sent to all nodes in the chassis to be throttled down to the lowest power level possible (CPU or Memory lowest P-state). In case the power usage remains too high, the chassis is shut down.
This is the default mode. This mode does not support the oversubscription mode (see “Power supply oversubscription” on page 54).
򐂰 N+1 Mode
One installed power supply is used as redundant. The failure of one power supply is allowed without affecting the system’s operation or performance (performance can be affected if oversubscription mode is enabled).
This mode can be enabled with oversubscription mode.
54 IBM NeXtScale System Planning and Implementation Guide
򐂰 N+N Mode
Half of the power supplies that are installed are used as redundant. The failure of up to half the number of the power supplies is allowed without affecting the system’s operation or performance (performance can be affected if oversubscription mode is enabled). This mode is useful if you have two power sources from two separate PDU for example.
This mode can be enabled with oversubscription mode.
3.8.4 Power supply oversubscription
By using oversubscription, users can make the most of the extra power from the redundant power supplies when the power supplies are in healthy condition.
For example, when oversubscription is enabled with N+1 redundancy mode, the total power that is available is equivalent to No Redundancy mode with six power supplies in the chassis. This means we can count on the power of six power supplies instead of five for normal operation.
When oversubscription mode is enabled with redundant power (N+1 or N+N redundancy), the total available power is 120% of the label ratings of the power supplies. So, for a 900 W power supply, it can be oversubscribed to 900 W x 120% = 1080 W.
For example, this means that with an N+1 power policy and six power supplies, instead of 5 x 900 W (4500 W) of power, you have 5 x 900 W x 120% (5400 W) of power that is available to the compute nodes.
Table 3-12 shows the power budget that is available, depending on the redundancy and oversubscription mode that is selected.
Table 3-12 Power budget for 6 x 900 W power supplies
Redundancy Mode Oversubscription mode Power budget
a
a. The power budget that is presented in this table is based on power supply
ratings. Actual power budget can vary.
Non-redundant Not available 5400 W (= 6 x 900 W)
N+1
Disabled 4500 W (= 5 x 900 W)
Enabled 5400 W (= 5 x 900 W x 120%)
N+N
Disabled 2700 W (= 3 x 900 W)
Enabled 3240 W (= 3 x 900 x 120%)
Chapter 3. IBM NeXtScale n1200 Enclosure 55
When oversubscription mode is enabled with redundant power (N+1 or N+N redundancy), the chassis’ total available power can be stretched beyond the label rating (up to 120%). However, the power supplies are designed to sustain this oversubscription for a limited time (approximately 1 second).
In healthy condition (all power supplies are in normal-operational mode), the redundant power supplies provide the extra 20% power oversubscription load for the rest of the normal-operational power supplies (none of the power supplies are oversubscribed).
When redundant power supplies fail (that is, one power supply failure in N+1 mode, or up to N power supplies fail in N+N mode), the remaining normal-operational power supplies provide the extra 20% power oversubscription load. This extra power is provided only for a limited time to allow the compute nodes to throttle to the lowest P-state to reduce their power usage back to a supported range. By design, the compute nodes perform this action quickly enough and operation continues.
The Table 3-13 shows the consequences of redundancy failure in the chassis with and without oversubscription mode.
Table 3-13 Consequences of power supply failure, depending on the oversubscription
Non-redundant mode: It is not possible to enable the oversubscription mode without any power redundancy.
Redundancy mode
Oversubscription mode
Consequences of redundancy failure
a
a. Considering one power supply failure in non-redundant and N+1 mode, and
three power supplies failures in N+N mode.
Compute nodes might be throttled
b
b. Compute nodes are throttled only if they require more power than what is
available on the remaining power supplies.
Chassis might power off
Non-redundant Not available Yes Yes
c
c. The chassis is powered off only if after throttling the compute nodes, the
enclosure power requirement still exceeds the power that is available on the remaining power supplies.
N+1
Disabled
No No
Enabled Yes
No
N+N
Disabled
No No
Enabled Yes
No
56 IBM NeXtScale System Planning and Implementation Guide
3.8.5 Acoustic mode
By using acoustic mode, the user has some control over the fan speeds and airflow (and noise) that is produced by the system fans. This mode can be used for noise or airflow concerns in the user environment.
When the option is set to Off, the fan speeds change as required for optimal cooling.
When the option is set to On, the chassis offers the following set points where the fan speeds are capped:
򐂰 1: highest acoustics attenuation (lowest cooling) 򐂰 2: intermediate acoustics attenuation 򐂰 3: low acoustics attenuation (higher cooling).
As a result, these settings increase the possibility that the node might have to be throttled to maintain cooling within the fan speed limitation.
The acoustic mode setting should be balanced with the customer’s workload for best performance.
Chapter 3. IBM NeXtScale n1200 Enclosure 57
3.9 Specifications
This section describes the specifications of the IBM NeXtScale n1200 Enclosure.
3.9.1 Physical specifications
The enclosure features the following physical specifications: 򐂰 Dimensions:
– Height: 263.3 mm (10.37 in.) – Depth: 914.5 mm (36 in.) – Width: 447 mm (17.6 in.)
򐂰 Weight:
– Fully configured (stand-alone): Approximately 112 kg (247 lbs.) – Empty chassis: approximately 28 kg (62 lbs.)
򐂰 Approximate heat output:
– Ship configuration: 341.18 Btu/hr (100 watts) – Full configuration: 20,470.84 Btu/hr (6,000 watts)
򐂰 Declared sound power level: 7.5 bels 򐂰 Chassis airflow:
Full chassis configuration with all compute nodes, FPC, power supplies, and fan modules installed:
– Minimum: 158 CFM (idle) – Maximum: 614 CFM
3.9.2 Supported environment
The IBM NeXtScale n1200 Enclosure complies with the following ASHRAE class A3 specifications.
򐂰 Power on
1
:
– Temperature: 5°C - 40°C (41°F - 104°F)
2
– Humidity, non-condensing: -12°C dew point (10.4°F) and 8% - 85%
relative humidity 5, 6 – Maximum dew point: 24°C (75°F) – Maximum altitude: 3048 m (10,000 ft.) – Maximum rate of temperature change: 5°C/hr. (41°F/hr.)
3
1
Chassis is powered on.
2
A3 - Derate maximum allowable temperature 1°C/175 m above 950 m.
58 IBM NeXtScale System Planning and Implementation Guide
򐂰 Power off4:
– Temperature: 5°C to 45°C (41°F - 113°F) – Relative humidity: 8% - 85% – Maximum dew point: 27°C (80.6°F)
򐂰 Storage (non-operating):
– Temperature: 1°C to 60°C (33.8°F - 140°F) – Altitude: 3050 m (10,006 ft.) – Relative humidity: 5% - 80% – Maximum dew point: 29°C (84.2°F)
򐂰 Shipment (non-operating)
5
:
– Temperature: -40°C to 60°C (-40°F - 140°F) – Altitude: 10700 m (35,105 ft.) – Relative humidity: 5% - 100% – Maximum dew point: 29°C (84.2°F)
6
3
5°C per hour for data centers that use tape drives and 20°C per hour for data centers that use disk drives.
4
Chassis is removed from original shipping container and is installed but not in use; for example, during repair, maintenance, or upgrade.
5
The equipment acclimation period is 1 hour per 20 °C of temperature change from the shipping environment to the operating environment.
6
Condensation is acceptable, but not rain.
© Copyright IBM Corp. 2013, 2014. All rights reserved. 59
Chapter 4. IBM NeXtScale nx360 M4
compute node
The IBM NeXtScale nx360 M4 compute node, machine type 5455, is a half-wide, dual-socket server that is designed for data centers that require high performance but are constrained by floor space. It supports Intel Xeon E5-2600 v2 series processors up to 12 cores, which provide more performance per server than previous generation systems.
With more computing power per watt and the latest Intel Xeon processors, you can reduce costs while maintaining speed and availability. A total of 12 nx360 M4 servers can be installed into the 6U NeXtScale n1200 enclosure.
This chapter describes the nx360 M4 compute node and includes the following topics:
򐂰 4.1, “Overview” on page 60 򐂰 4.2, “System architecture” on page 63 򐂰 4.3, “Specificiations” on page 67 򐂰 4.4, “Standard models” on page 69 򐂰 4.5, “Processor options” on page 70 򐂰 4.6, “Memory options” on page 71 򐂰 4.7, “Internal disk storage options” on page 77 򐂰 4.8, “IBM NeXtScale Storage Native Expansion Tray” on page 86 򐂰 4.9, “IBM NeXtScale PCIe Native Expansion Tray” on page 90
4
60 IBM NeXtScale System Planning and Implementation Guide
򐂰 4.10, “GPU and coprocessor adapters” on page 92 򐂰 4.11, “Embedded 1 Gb Ethernet controller” on page 96 򐂰 4.12, “PCI Express I/O adapters” on page 97 򐂰 4.13, “Integrated virtualization” on page 102 򐂰 4.14, “Local server management” on page 103 򐂰 4.15, “Remote server management” on page 105 򐂰 4.16, “External disk storage expansion” on page 107 򐂰 4.17, “Physical specifications” on page 110 򐂰 4.18, “Operating systems support” on page 112
4.1 Overview
The IBM NeXtScale nx360 M4 compute node contains only the essential components in the base architecture to provide a cost-optimized platform. The nx360 M4 compute node provides a dense, flexible solution with a low total cost of ownership.
Figure 4-1 IBM NeXtScale nx360 M4 compute node
The nx360 M4 compute nodes fit into the IBM NeXtScale n1200 Enclosure that provides common power and cooling resources. As shown in Figure 4-1, the nx360 M4 compute node can support the following components:
򐂰 One or two Intel Xeon E5-2600 v2 series processors. 򐂰 Up to eight DIMMs of registered DDR3 ECC memory and operating up to
1866 MHz, which provides a total memory capacity of up to 128 GB.
򐂰 One on-board 1 Gb Ethernet port and one on-board 1 Gb
Ethernet/management port.
Chapter 4. IBM NeXtScale nx360 M4 compute node 61
򐂰 An Integrated Management Module II (IMM2) port for server remote
management and an integrated industry-standard Unified Extensible Firmware Interface (UEFI), which enables improved setup, configuration, and updates.
򐂰 One 3.5-inch drive bay, or alternatively two 2.5-inch drive bays or four
1.8-inch drive bays.
򐂰 Support for additional local storage with the use of the Storage Native
Expansion Tray. When using 4 TB HDDs, you can create an ultra-dense storage server with up to 32 TB of total disk capacity within 1U of comparable rack density.
򐂰 A slot for 10 Gb Ethernet or FDR InfiniBand mezzanine for network
connectivity without using a PCIe slot.
򐂰 PCI Express 3.0 I/O expansion capabilities through a 16x riser cage.
Additional support for two GPUs or coprocessors with the use of the PCIe Native Expansion Tray.
4.1.1 Physical design
Figure 4-2 shows the controls and connections on the front of the server.
Figure 4-2 Front view of IBM NeXtScale nx360 M4
Power
button and
information
LEDs
PCIe 3.0 slot
full height, half length
Dual-port
mezzanine
adapter slot
KVM
connector
1 GbE
(dedicated)
1 GbE or IMM
management port (shared)
IMM
management
port (dedicated)
62 IBM NeXtScale System Planning and Implementation Guide
Figure 4-3 shows the locations of key components inside the server.
Figure 4-3 Inside view of the IBM NeXtScale nx360 M4
CPU 2 and
four DIMMs
CPU 1 and
four DIMMs
Mezz card
connector
PCIe 3.0
riser slot 1
PCIe 3.0
riser slot 2
(PCIe Tray only)
USB hypervisor
socket
Midplane
connector
KVM and
Ethernet
Drive bay(s)
SATA port and cable
Chapter 4. IBM NeXtScale nx360 M4 compute node 63
Figure 4-4 shows an exploded window of the platform, in which the major components and options are highlighted.
Figure 4-4 Exploded view of the nx360 M4
4.2 System architecture
The IBM NeXtScale nx360 M4 compute node features the Intel E5-2600 v2 series processors. The Xeon E5-2600 v2 series processor has models with either 6, 8, 10, or 12 cores per processor with up to 24 threads per socket.
The Intel Xeon E5-2600 v2 series processors (formerly known by the Intel code name
Ivy Bridge-EP) are the successors of the first implementation of Intel’s
micro architecture that is based on tri-gate transistors, Intel Xeon E5-2600 series (formerly
Sandy Bridge-EP). The Xeon E5-2600 v2 series uses a 22nm
manufacturing process in contrast to the 32nm used by its predecessor. By using this new, smaller 22nm tri-gate transistor technology, you can design more powerful processors with better power efficiency.
Cover
1.8-inch drives, backplane & cage
2.5 drives, backplane & cage
3.5-inch drive and carrier
DIMM
Air baffle
PCIe riser
Battery
holder
Air baffle
Slot
covers
Air
baffles
64 IBM NeXtScale System Planning and Implementation Guide
Such new tri-gate transistor technology enabled a new architecture with which you can share data on-chip through a high-speed ring that is interconnected between all processor cores, the last level cache (LLC), and the system agent. The system agent houses the memory controller and a PCI Express root complex that provides 40 PCIe 3.0 lanes.
The integrated memory controller in each CPU still supports four memory channels with three DDR3 DIMMs per channel but now runs at a speed that is up to 1866 MHz. Two QPI links still also connect to a second CPU in a dual-socket installation.
The Xeon E5-2600 v2 series is available with up to 12 cores and 30 MB of last-level cache. It features an enhanced instruction set that is called Intel Advanced Vector Extensions (AVX). It doubles the operand size for vector instructions (such as floating-point) to 256 bits and boosts selected applications by up to a factor of two.
The implementation architecture includes Intel Turbo Boost Technology 2.0 and improved power management capabilities. Turbo Boost automatically turns off unused processor cores and increases the clock speed of the cores in use if thermal requirements are still met. Turbo Boost Technology 2.0 uses the integrated design and implements a more granular overclocking in 100 MHz steps instead of 133 MHz steps on older microprocessors.
As with iDataPlex servers, NeXtScale servers support S3 mode. S3 allows systems to come back into full production from low-power state much quicker than a traditional power-on. In fact, cold boot normally takes about 270 seconds; with S3, cold boot occurs in only 45 seconds. When you know a system will not be used because of time of day or state of job flow, you can send it into a very low-power state to save power and bring it back online quickly when needed.
Table 4-1 summarizes the differences between both Intel’s micro architecture implementations. Improvements are highlighted.
Table 4-1 Comparison between Xeon E5-2600 and Xeon E5-2600 v2
Xeon E5-2600
(Sandy Bridge-EP)
Xeon E5-2600 v2
(Ivy Bridge-EP)
QPI Speed (GT/s) 8.0, 7.2 and 6.4 GT/s
Addressability 46 bits physical, 48 bits virtual
Cores Up to 8
Up to 12
Threads per socket Up to 16 threads
Up to 24 threads
Last-level Cache (LLC) Up to 20 MB
Up to 30 MB
Chapter 4. IBM NeXtScale nx360 M4 compute node 65
Figure 4-5 shows the IBM NeXtScale nx360 M4 building block.
Figure 4-5 IBM NeXtScale nx360 M4 system board block diagram
Intel Turbo Boost Technology Yes
Memory population 4 channels of up to 3 RDIMMs, 3 LRDIMMs, or 2 UDIMMs
Maximum memory speed Up to 1600 MHz Up to 1866 MHz
Memory RAS features ECC, Patrol Scrubbing, Sparring, Mirroring, Lockstep Mode, x4/x8
SDDC
PCIe lanes 40 PCIe 3.0 lanes
TDP values (W) 130, 115, 96, 80, 70, 60, 50 W
Idle power targets (W) 15 W or higher
12 W for low-voltage SKUs
10.5 W or higher
7.5 W for low-voltage SKUs
Xeon E5-2600
(Sandy Bridge-EP)
Xeon E5-2600 v2
(Ivy Bridge-EP)
DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
QPI links
x4 ESI link
PCIe riser slot
1x 3.5” or 2x 2.5” or 4x 1.8”
2x 6 Gbps AHCI & 2x 3 Gbps AHCI
USB
Internal
Intel
C602J
PCH
Dedicated RJ45 port
Front KVM port
IMM v2
Video & serial
Mezzanine slot
GPU riser slot
PCIe 3.0 x24
PCIe 3.0 x8
PCIe 3.0 x16
Intel
Xeon
Processor 1
DDR3 DIMMs
4 memory channels
1 DIMM per channel
x1 USB
Intel
Xeon
Processor 2
USB
PCIe 3.0 x4
GbE
Two GbE ports (RJ45)
66 IBM NeXtScale System Planning and Implementation Guide
The nx360 M4 architecture contains the following components:
򐂰 Two 2011-pin, Socket R (LGA-2011) processor sockets 򐂰 Intel C602J Platform Controller Hub (PCH) 򐂰 Four memory channels per socket 򐂰 One DIMM per memory channel (DPC) 򐂰 Eight DDR3 DIMM sockets (when only one processor is installed, only four
DIMM sockets can be used)
򐂰 Support for UDIMMs and RDIMMs 򐂰 A PCI Express 3.0 x8 slot for Mezzanine card connection that connects to
CPU 1
򐂰 A PCI Express 3.0 x24 slot for PCIe riser cage that provides a
full-height/half-length (FHHL) slot that connects to CPU 1
򐂰 A PCI Express 3.0 x16 slot that is connected to CPU 2 (for use with attached
trays)
򐂰 Dual-port integrated 1 Gb controller 򐂰 An Integrated Management Module v2 (IMMv2) for remote server
Note: Although the Socket R (LGA-2011) processor socket on the nx360 M4 physically fits Xeon E5-2600 series and Xeon E5-2600 v2 series processor, only the latter are supported as processor options for this platform.
Chapter 4. IBM NeXtScale nx360 M4 compute node 67
4.3 Specificiations
Table 4-2 lists the standard specifications of the NeXtScale nx360 M4 compute node and NeXtScale n1200 enclosure.
Table 4-2 Specificaitons
Components Specification
Form factor Half-wide, 1U compute node
Supported chassis
IBM NeXtScale n1200 enclosure, 6U high; up to 12 compute nodes per chassis
Processor Two Intel Xeon Processor E5-2600 V2 series processors; QuickPath Interconnect
(QPI) links speed up to 8.0 GT/s. Hyper-Threading Technology and Turbo Boost Technology. Intel C602J (Patsburg-J) chipset. Processor options:
򐂰 4-core processors up to 3.5 GHz and 15 MB L3 cache 򐂰 6-core processors up to 3.5 GHz and 25 MB L3 cache 򐂰 8-core processors up to 3.3 GHz and 25 MB L3 cache 򐂰 10-core processors up to 3.0 GHz and 25 MB L3 cache 򐂰 12-core processors up to 2.7 GHz and 30 MB L3 cache
Memory Up to 8 DIMM sockets (4 DIMMs per processor) supporting DDR3 DIMMs up to 1866
MHz memory speeds. RDIMMs, UDIMMs and LRDIMMs supported. Four memory channels per processor (one DIMM per channel).
Memory maximums
Up to 256 GB with 8x 32 GB LRDIMMs and two processors.
Memory protection
ECC, memory mirroring, and memory sparing.
Disk drive bays Inside the nx360 M4: One 3.5-inch simple-swap SATA or two 2.5-inch simple swap
SAS/SATA HDDs or SSDs, or four 1.8-inch simple-swap SSDs. Not front accessible. Adding the NeXtScale Storage Native Expansion Tray adds 7 more 3.5-inch simple-swap drive bays (adds 1U height).
Maximum internal storage
With the Storage Native Expansion Tray: 32 TB using 8x 4TB 3.5-inch drives. Without the Storage Native Expansion Tray: 4.0 TB using 1x 4TB 3.5-inch drive.
RAID support On some models: ServeRAID C100 6Gb SATA controller supporting RAID-0, RAID-1
and RAID-10. Implemented in the Intel C600 chipset. Optional hardware RAID with supported 6Gbps RAID controllers.
Optical drive bays
No internal bays. Use an external USB drive such as the IBM and Lenovo part number 73P4515 or 73P4516.
Tape drive bays No internal bays. Use an external USB drive.
68 IBM NeXtScale System Planning and Implementation Guide
Network interfaces
Two Gigabit Ethernet ports using onboard Intel I350 Gb Ethernet controller. Optionally, two InfiniBand ports or two 10 GbE ports via a mezzanine card (which does not occupy the available PCIe slot).
PCI Expansion slots
nx360 M4 without PCIe Native Expansion Tray:
򐂰 One PCIe 3.0 x8 mezzanine card slot 򐂰 One PCIe 3.0 x16 full-height half-length slot
nx360 M4 with PCIe Native Expansion Tray (addes 1U height):
򐂰 One PCIe 3.0 x8 mezzanine card slot 򐂰 One PCIe 3.0 x8 full-height half-length slot 򐂰 Two PCIe 3.0 x16 full-height full-length double-width slots for GPUs
Ports (server) Front of the server: KVM connector; with the addition of a console breakout cable (1
cable standard with the chassis) supplies one RS232 serial port, one VGA port and two USB ports for local console connectivity. Three 1 Gbps Ethernet ports with RJ45 connectors: one dedicated for systems management (wired to the IMM), one dedicated for use by the operating system and one shared by the IMM and the operating system. One slot for an optional mezzanine card ports (QSFP, SFP+ or RJ45 depending on the card installed). One internal USB port for VMware ESXi hypervisor key.
Ports (chassis) Rear of the enclosure, provided by the Fan and Power Controller Module for chassis
management: Gb Ethernet connection (RJ45) for browser-based remote management, mini-USB serial port for local management.
Cooling Supplied by the NeXtScale n1200 enclosure. 10 hot-swap dual-rotor 80 mm system
fans with tool-less design.
Power supply Supplied by the NeXtScale n1200 enclosure. Up to six hot-swap power supplies either
900W or 1300W depending on the chassis model. Support power policies N+N or N+1 power redundancy. 80 PLUS Platinum certified.
Video Matrox G200eR2 video core with 16 MB DDR3 video memory integrated into the
IMM2. Maximum resolution is 1600x1200 with 16M colors (32 bpp) at 75 Hz, or 1680x1050 with 16M colors at 60 Hz. Optional GPUs in PCIe Native Exp. Tray.
Systems management
UEFI, IBM Integrated Management Module II (IMM2) with Renesas SH7757 controller, Predictive Failure Analysis, Light Path Diagnostics, Automatic Server Restart, IBM Systems Director and IBM Systems Director Active Energy Manager, IBM ServerGuide. Browser-based chassis management via Ethernet port on the Fan and Power Controller Module on the rear of the enclosure. IMM2 upgrades available to IMM2 Standard and IMM2 Advanced for web GUI and remote presence features.
Security features Power-on password, administrator's password, Trusted Platform Module 1.2.
Operating systems supported
Red Hat Enterprise Linux, SUSE Linux Enterprise Server, Microsoft Windows Server 2008 R2 and 2008, VMware vSphere Hypervisor.
Components Specification
Chapter 4. IBM NeXtScale nx360 M4 compute node 69
4.4 Standard models
Table 4-3 lists the standard models of nx360 M4.
Table 4-3 Standard models of the nx360 M4
Limited warranty 3-year customer-replaceable unit and onsite limited warranty with 9x5/NBD.
Service and support
Optional service upgrades are available through IBM ServicePacs®: 4-hour or 2-hour response time, 8-hour fix time, 1-year or 2-year warranty extension, remote technical support for IBM hardware and some IBM and OEM software.
Dimensions NeXtScale nx360 M4 server: Width: 216 mm (8.5 in), height: 41 mm (1.6 in), depth:
659 mm (25.9 in) NeXtScale n1200 enclosure: Width: 447 mm (17.6 in), height: 263 mm (10.4 in), depth: 915 mm (36 in.)
Weight NeXtScale nx360 M4 maximum weight: 6.05 kg (13.31 lb)
NeXtScale n1200 enclosure: Fully configured (stand-alone): 112 kg (247 lb), empty chassis 28 kg (62 lb)
Components Specification
Model Intel Xeon processor
a
a. Processor detail: Processor quantity and model, cores, core speed, L3 cache, memory speed, and power consumption.
Memory and speed
Disk adapter
Disk bays
Disks Network
5455-22x 2x Intel Xeon E5-2620 v2 6C
2.1GHz 15MB 1600MHz 80W
2x 4 GB 1600 MHz
6 Gbps SATA (No RAID)
1x 3.5” SS bay
b
b. SS = simple swap
Open 2x GbE
5455-42x 2x Intel XeonE5-2660 v2 10C
2.2GHz 25MB 1866MHz 95W
2x 8 GB 1866 MHz
6 Gbps SATA (No RAID)
1x 3.5” SS bay
Open 2x GbE
5455-62x 2x Intel XeonE5-2670 v2 10C
2.5GHz 25MB 1866MHz 115W
2x 8 GB 1866 MHz
ServeRAID C100
2x 2.5" SS bays
Open 2x GbE
70 IBM NeXtScale System Planning and Implementation Guide
4.5 Processor options
The nx360 M4 supports Xeon E5-2600 v2 processor series. The exact processor options that can be selected for nx360 M4 compute node are listed in Table 4-4.
Table 4-4 Processor options
Part number
Feature code
a
a. The first feature code corresponds to the first processor; the second feature code corresponds to
the second processor.
Intel Xeon processors
b
Where used
00FL128 A55N / A55W Intel Xeon E5-2603 v2 4C 1.8GHz 10MB 1333MHz 80W -
00FL129 A55P / A55X Intel Xeon E5-2609 v2 4C 2.5GHz 10MB 1333MHz 80W -
00Y8687 A4MF / A4MK Intel Xeon E5-2618L v2 6C 2.0GHz 15MB 1333MHz 50W -
46W2712 A425 / A42F Intel Xeon E5-2620 v2 6C 2.1GHz 15MB 1600MHz 80W 22x
00FL130 A55Q / A55Y Intel Xeon E5-2628L v2 8C 2.2GHz 20MB 1600MHz 70W -
00FL234 A55T / A561 Intel Xeon E5-2630 v2 6C 2.6GHz 15M 1600MHz 80W -
00FL131 A55R / A55Z Intel Xeon E5-2630L v2 6C 2.4GHz 15MB 1600MHz 60W -
00Y8632 A4MD / A4MH Intel Xeon E5-2637 v2 4C 3.5GHz 15MB 1866MHz 130W -
46W2719 A42B / A42M Intel Xeon E5-2640 v2 8C 2.0GHz 20MB 1600MHz 95W -
00FL126 A55L / A55U Intel Xeon E5-2643 v2 6C 3.5GHz 25MB 1866MHz 130W -
00Y8686 A4ME / A4MJ Intel Xeon E5-2648L v2 10C 2.0GHz 25MB 1866MHz 70W -
46W2713 A426 / A42G Intel Xeon E5-2650 v2 8C 2.6GHz 20MB 1866MHz 95W -
00FL132 A55S / A560 Intel Xeon E5-2650L v2 10C 1.7GHz 25M 1600MHz 70W -
00Y8688 A4MG / A4ML Intel Xeon E5-2658 v2 10C 2.4GHz 25MB 1866MHz 95W -
46W2714 A427 / A42H Intel Xeon E5-2660 v2 10C 2.2GHz 25MB 1866MHz 95W 42x
00FL127 A55M / A55V Intel Xeon E5-2667 v2 8C 3.3GHz 25MB 1866MHz 130W -
46W2715 A428 / A42J Intel Xeon E5-2670 v2 10C 2.5GHz 25MB 1866MHz 115W 62x
46W2716 A429 / A42K Intel Xeon E5-2680 v2 10C 2.8GHz 25MB 1866MHz 115W -
46W2717 A42A / A42L Intel Xeon E5-2690 v2 10C 3.0GHz 25MB 1866MHz 130W -
46W2720 A42C / A42N Intel Xeon E5-2695 v2 12C 2.4GHz 30MB 1866MHz 115W -
46W2721 A42D / A42P Intel Xeon E5-2697 v2 12C 2.7GHz 30MB 1866MHz 130W -
Chapter 4. IBM NeXtScale nx360 M4 compute node 71
4.6 Memory options
IBM DDR3 memory is compatibility tested and tuned for optimal IBM System x performance and throughput. IBM memory specifications are integrated into the light path diagnostic tests for immediate system performance feedback and optimum system uptime. From a service and support standpoint, IBM memory automatically assumes the IBM system warranty, and IBM provides service and support worldwide.
The NeXtScale nx360 M4 supports DDR3 memory. The server supports up to four DIMMs when one processor is installed and up to eight DIMMs when two processors are installed. Each processor has four memory channels. There is one DIMM per memory channel (1 DPC).
Figure 4-6 shows the memory channel layout.
Figure 4-6 Memory channel layout of the IBM NeXtScale nx360 M4
b. Processor detail: Model, cores, core speed, L3 cache, memory speed, and TDP power.
Floating point performance: The number of sockets and the processor option that are selected determine the theoretical floating point peak performance, as shown in the following example:
#sockets x #cores per processor x freq x 8 flops per cycle = Gflops
A nx360 compute node with dual socket E5-2680 v2 series 10-core that operates at 2.8 Ghz has the following peak performance:
2 x 10 x 2.8 x 8 = 448 Gflops
DIMM 3DIMM 2DIMM 1
Xeon E5-2600 v2
Processor 1
Channel A
Memory Controller
Channel B Channel C
QPI links
DIMM 4
Channel D
DIMM 7DIMM 6DIMM 5
Xeon E5-2600 v2
Processor 2
Channel A
Memory Controller
Channel B Channel C
DIMM 8
Channel D
72 IBM NeXtScale System Planning and Implementation Guide
The following rules apply when you select the memory configuration:
򐂰 Each installed processor must have at least one memory DIMM connected. 򐂰 In the nx360 M4, the maximum memory speed of a configuration is the lower
of the following two values:
– The memory speed of the processor (see Table 4-4 on page 70) – The memory speed of the DIMM (see Table 4-10 on page 76)
򐂰 The server supports 1.5 V and 1.35 V DIMMs. Mixing 1.5 V and 1.35 V
DIMMs in the same server is supported. In such a case, all DIMMs operate at
1.5 V.
򐂰 Mixing UDIMMs and RDIMMs is not supported. 򐂰 Equally distribute DIMMs between sockets for best performance. 򐂰 For optimal performance, populate the four memory channels when one
processor is installed and the eight DIMMs when two processors are installed.
Table 4-5 lists the memory options that are available for the nx360 M4 server.
Table 4-5 Table 5. Memory options
Part number
Feature code
Description Maximum
supported
Models where used
UDIMMs
00D5012 A3QB 4GB (1x4 GB, 2Rx8, 1.35 V) PC3L-12800
CL11 ECC DDR3 1600MHz LP UDIMM
8-
RDIMMs - 1866 MHz
00D5028 A3QF 4GB (1x4 GB, 2Rx8, 1.5 V) PC3-14900
CL13 ECC DDR3 1866MHz LP RDIMM
8-
00D5040 A3QJ 8GB (1x8 GB, 2Rx8, 1.5 V) PC3-14900
CL13 ECC DDR3 1866MHz LP RDIMM
8 42x, 62x
00D5048 A3QL 16GB (1x16 GB, 2Rx4, 1.5 V) PC3-14900
CL13 ECC DDR3 1866MHz LP RDIMM
8-
RDIMMs - 1600 MHz
00D5024 A3QE 4GB (1x4GB, 1Rx4, 1.35V) PC3L-12800
CL11 ECC DDR3 1600MHz LP RDIMM
8-
46W0735 A3ZD 4GB (1x4 GB, 2Rx8, 1.35 V) PC3-12800
CL13 ECC DDR3 1600MHz LP RDIMM
822x
Chapter 4. IBM NeXtScale nx360 M4 compute node 73
The following memory protection technologies are supported:
򐂰 ECC 򐂰 Chipkill (for x4-based memory DIMMs; look for “x4” in the DIMM description;
for x8-based memory DIMMs, only ECC protection is supported)
򐂰 Memory mirroring mode 򐂰 Memory lock-step mode
When mirrored mode is used, DIMM 1 and DIMM 2 hold the same data for redundancy purposes. If one DIMM fails, it is disabled and the backup DIMM in the other channel takes over. Likewise, DIMM 3 and DIMM 4 hold the same data and create a mirrored pair. Because memory mirroring is handled in hardware, it is operating system-independent. The total usable memory size is half the total of the installed memory.
Lock-step mode requires paired memory channels (DIMM 1 and DIMM 2, DIMM 3 and DIMM 4) to be populated by the same memory regards the size and the organization. Lock-step mode allows Single Device Data Correction (SDDC) memory protection for x8-based memory DIMMs.
4.6.1 DIMM installation order
The IBM NeXtScale nx360 M4 boots with only one memory DIMM installed per processor. However, the suggested memory configuration is to balance the memory across all the memory channels on each processor to use the available memory bandwidth. For best performance, it is recommended to populate all DIMM slots.
00D5036 A3QH 8GB (1x8GB, 1Rx4, 1.35V) PC3L-12800
CL11 ECC DDR3 1600MHz LP RDIMM
8-
00D5044 A3QK 8GB (1x8 GB, 2Rx8, 1.35 V) PC3L-12800
CL11 ECC DDR3 1600MHz LP RDIMM
8-
46W0672 A3QM 16GB (1x16GB, 2Rx4, 1.35V) PC3L-12800
CL11 ECC DDR3 1600MHz LP RDIMM
8-
LRDIMMs
46W0761 A47K 32GB (1x32GB, 4Rx4, 1.5V) PC3-14900
CL13 ECC DDR3 1866MHz LP LRDIMM
8-
Part number
Feature code
Description Maximum
supported
Models where used
74 IBM NeXtScale System Planning and Implementation Guide
The locations of the DIMM sockets relative to the processors are shown in Figure 4-7.
Figure 4-7 Memory DIMM slots
Memory DIMM installation: Independent channel mode
Independent channel mode provides a maximum memory of 64 GB of usable memory with one installed processor, and 128 GB of usable memory with two installed microprocessor (using 16 GB DIMMs).
Table 4-6 shows DIMM installation if you have one processor that is installed.
Table 4-6 Memory population with one processor installed
Table 4-7 on page 75 shows DIMM installation if you have two processors that are installed. A minimum of two memory DIMMs (one for each processor) are required when two processors are installed.
DIMM 1DIMM 2
DIMM 3
DIMM 4
DIMM 8DIMM 7
CPU 1
CPU 2
DIMM 5DIMM 6
Processor 1
Number of DIMMs DIMM 1 DIMM 2 DIMM 3 DIMM 4
1 x
2
x x
3
x x x
4
x x x x
Chapter 4. IBM NeXtScale nx360 M4 compute node 75
Table 4-7 Memory population table with two processors installed
Memory DIMM installation: Mirrored-channel mode
In mirrored channel mode, the channels are paired and both channels in a pair store the same data.
As shown in Table 4-8, for each microprocessor, DIMM 1 and DIMM 2 form a redundant pair, and DIMM 3 and DIMM 4 form the other redundant pair. Because of the redundancy, the effective memory capacity of the compute node is half the installed memory capacity.
The pair of DIMMs that are installed in each channel must be identical in capacity, type, and rank count.
Table 4-8 Memory population with processor stalled: Mirrored channel mode
Table 4-9 on page 76 shows DIMM installation for memory mirroring if two processors are installed.
Processor 1 Processor 2
Number of DIMMs
DIMM 1 DIMM 2 DIMM 3 DIMM 4 DIMM 5 DIMM 6 DIMM 7 DIMM 8
2
x x
3
x x x
4
x x x x
5
x x x x x
6
x x x x x x
7
x x x x x x x
8
x x x x x x x x
Processor 1
Number of DIMMs DIMM 1 DIMM 2 DIMM 3 DIMM 4
2 x x
4
x x x x
76 IBM NeXtScale System Planning and Implementation Guide
Table 4-9 Memory population table with two processors installed: Mirrored channel mode
Table 4-10 (RDIMMs) and Table 4-11 on page 77 (UDIMMs and LRDIMMs) show the maximum memory speeds that are achievable. The tables also show the maximum memory capacity at any speed that is supported by the DIMM and the maximum memory capacity at the rated DIMM speed.
Table 4-10 Maximum memory speeds (RDIMMs)
Processor 1 Processor 2
Number of
DIMMs
DIMM 1 DIMM 2 DIMM 3 DIMM 4 DIMM 5 DIMM 6 DIMM 7 DIMM 8
4 x x x x
6
x x x x x x
8
x x x x x x x x
Spec RDIMMs
Rank Single rank Dual rank
Part numbers 00D5024 (4GB)
00D5036 (8GB)
46W0735 (4 GB)
00D5044 (8 GB)
46W0672 (16GB)
00D5028 (4 GB) 00D5040 (8 GB)
00D5048 (16 GB)
Rated speed 1600 MHz 1600 MHz 1866 MHz
Rated voltage 1.35V 1.35 V 1.5 V
Operating voltage 1.35V or 1.5V 1.35V or 1.5V 1.5 V
Max quantity
a
a. The maximum quantity that is supported is shown for two installed processors. When one processor
is installed, the maximum quantity that is supported is half of that shown.
88 8
Largest DIMM 8 GB 16 GB 16 GB
Max memory capacity 64 GB 128 GB 128 GB
Max memory at rated speed 64 GB 128 GB 128 GB
Maximum operating speed (MHz)
1 DIMM per channel 1600 MHz 1600 MHz 1866 MHz
Chapter 4. IBM NeXtScale nx360 M4 compute node 77
Table 4-11 Maximum memory speeds (UDIMMs and LRDIMMs)
4.7 Internal disk storage options
This section provides information about the internal storage options. The RAID controller cards and the conventional and solid-state disks (SDDs) that are supported also are listed.
The IBM NeXtScale nx360 M4 server supports one of the following drive options:
򐂰 One 3.5-inch simple-swap SATA drive 򐂰 Up to two 2.5-inch simple-swap SAS or SATA HDDs or SSDs 򐂰 Up to four 1.8-inch simple-swap SSDs
Rules for mixing drive types: 򐂰 Mixing HDDs: Simple-swap SATA HDDs and simple-swap SAS HDDs can be
intermixed in the system, but cannot be intermixed in the same RAID array.
򐂰 Mixing HDDs and SSDs: Both simple-swap SATA HDDs and simple-swap
SAS HDDs can be intermixed with SSDs in the system. SAS or SATA HDDs cannot be configured with SSDs within the same RAID array.
Spec UDIMMs LRDIMMs
Rank Dual rank Quad rank
Part numbers 00D5012 (4GB) 46W0761 (32GB)
Rated speed 1600 MHz 1866 MHz
Rated voltage 1.35V 1.5 V
Operating voltage 1.35V or 1.5V 1.5 V
Max quantity
a
a. The maximum quantity that is supported is shown for two installed processors. When one processor
is installed, the maximum quantity that is supported is half of that shown.
88
Largest DIMM 4 GB 32 GB
Max memory capacity 32 GB 256 GB
Max memory at rated speed 32 GB 256 GB
Maximum operating speed (MHz)
1 DIMM per channel 1600 MHz 1866 MHz
78 IBM NeXtScale System Planning and Implementation Guide
In addition, the nx360 M4 supports seven additional 3.5-inch drive bays if the IBM NeXtScale Storage Native Expansion Tray is attached. The Storage Native Expansion Tray can be used with any of the above three internal drive configurations to provide the following bay combinations:
򐂰 Eight 3.5-inch simple-swap SATA drives 򐂰 Seven 3.5-inch simple-swap SATA drives and two 2.5-inch simple-swap SATA
HDDs
򐂰 Seven 3.5-inch simple-swap SATA drives and four 1.8-inch simple-swap
SSDs
We describe the Native Expansion Tray in detail in 4.8, “IBM NeXtScale Storage Native Expansion Tray” on page 86.
Drive cages for the drives internal to the nx360 M4 are as listed in Table 4-12. Drives used in the Storage Native Expansion Tray do not need a cage.
Table 4-12 Drive cages for the drive bay in the nx360 M4
Figure 4-8 shows the three disk drive bay options that are available for nx360 M4 compute node.
Figure 4-8 Drive bay options
Part number Feature
code
Description Models
where used
None
a
a. CTO only
A41N nx360 M4 1.8-inch SSD Cage Assembly -
None
a
A41K nx360 M4 2.5-inch HDD Cage Assembly 62x
None
a
A41J nx360 M4 3.5-inch HDD Cage Assembly 22x, 42x
00Y8615 A4GE 3.5" HDD RAID cage for nx360 M4
Storage Native Expansion Tray
-
O e35 c a dds d e
p p
One 3.5-inch HDD Up to two 2.5-inch HDDs or SSDs Up to four 1.8-inch SSD
Chapter 4. IBM NeXtScale nx360 M4 compute node 79
The 3.5-inch drive slides in place. The 2.5-inch and 1.8-inch drive bays pivot up so that you can insert the drives. You then pull the release pin and lower the drive bays back into place.
There are two 3.5-inch drive cages - the last two rows of Table 4-12 on page 78. If the Storage Native Expansion Tray is attached to the nx360 M4, then the use of the RAID cage (feature A4GE, option 00Y8615) allows you to configure a RAID array that spans all 8 drives - the 7 in the storage tray and the 1 drive internal to the nx360 M4. Such a configuration would be connected either to a ServeRAID M1115 adapter or N2115 SAS HBA.
If the 3.5-inch HDD cage (feature A41J) is used, then a RAID array can only be formed with the 7 drives in the storage tray. In such a configuration, the drives in the storage tray are connected either to a ServeRAID M1115 adapter or N2115 SAS HBA, and the single drive in the nx360 M4 is connected to the ServeRAID C100.
4.7.1 Controllers for internal storage
The nx360 M4 server support the following disk controllers: 򐂰 ServeRAID C100: An onboard SATA controller with software RAID
capabilities
򐂰 ServeRAID H1110 SAS/SATA: An entry-level hardware RAID controller that
integrates popular SAS technology
򐂰 ServeRAID M1115 SAS/SATA: An advanced RAID controller with cache/flash
modules and energy packs, and software feature upgrades in an flexible offerings structure.
򐂰 N2115 SAS/SATA HBA for IBM System x: A high-performance host bus
adapter for internal drive connectivity.
Table 4-13 lists the ordering information for RAID controllers and SAS HBA.
Table 4-13 Table 8. Controllers for internal storage
Part number Feature code Description
RAID controllers
None A17T ServeRAID C100 for System x
a
81Y4492 A1XL ServeRAID H1110 SAS/SATA Controller for IBM System x
81Y4448 A1MZ ServeRAID M1115 SAS/SATA Controller for IBM System x
46C8988 A3MW N2115 SAS/SATA HBA for IBM System x
80 IBM NeXtScale System Planning and Implementation Guide
ServeRAID C100 controller
The ServeRAID C100 is an integrated SATA controller with software RAID capabilities. It is a cost-effective way to provide reliability, performance, and fault-tolerant disk subsystem management.
The ServeRAID C100 has the following specifications:
򐂰 Supports RAID levels 0, 1, and 10 򐂰 Onboard SATA controller with software RAID capabilities 򐂰 Supports SATA HDDs and SATA SSDs 򐂰 Offers two 6-Gbps SATA ports and two 3-Gbps SATA ports 򐂰 Support for up to two virtual drives 򐂰 Support for virtual drive sizes greater than 2 TB 򐂰 Fixed stripe unit size of 64 KB 򐂰 Support for MegaRAID Storage Manager management software
For more information, see the list of IBM Redbooks Product Guides in the RAID adapters category at this website:
http://www.redbooks.ibm.com/portals/systemx?Open&page=pg&cat=raid
ServeRAID H1110 SAS/SATA controller
The ServeRAID H1110 SAS/SATA Controller for IBM System x offers a low-cost, enterprise-grade RAID solution for internal HDDs. It features a PCI Express 2.0 x4 host interface, MD0 form factor, and robust hardware RAID processing engine that is based on the LSI SAS2004 RAID on Chip (ROC) controller.
The ServeRAID H1110 adapter has the following specifications:
򐂰 Four internal 6 Gbps SAS/SATA ports 򐂰 One x4 mini-SAS internal connector (SFF-8087) 򐂰 6 Gbps throughput per port 򐂰 Based on LSI SAS2004 6 Gbps RAID on Chip (ROC) controller 򐂰 PCIe 2.0 x4 host interface
Upgrades
81Y4542 A1X1 ServeRAID M1100 Series Zero Cache/RAID 5 Upgrade (for M1115 only)
a. Windows and Linux only. No support for VMware, Hyper-V, or Xen
Part number Feature code Description
Note: The ServeRAID C100 is supported by Windows and Linux only.
Depending on the operating system version, drivers might need to be downloaded separately. There is no support for VMware, Hyper-V, or Xen
Chapter 4. IBM NeXtScale nx360 M4 compute node 81
򐂰 Supports RAID 0, 1, 1E, and 10 򐂰 SAS and SATA drives are supported, but the mixing of SAS and SATA in the
same integrated volume is not supported
򐂰 Supports up to two integrated volumes 򐂰 Supports up to two global hot-spare drives 򐂰 Supports drive sizes greater than 2 TB for RAID 0, 1E, and 10 (not RAID 1) 򐂰 Fixed stripe size of 64 KB
For more information, see the list of IBM Redbooks Product Guides in the RAID adapters category at this website:
http://www.redbooks.ibm.com/portals/systemx?Open&page=pg&cat=raid
ServeRAID M1115 controller
The ServeRAID M1115 SAS/SATA Controller is a part of the IBM ServeRAID M Series family that offers a complete server storage solution, which consists of RAID controllers, cache/flash modules, energy packs, and software feature upgrades in an ultra-flexible offerings structure. M1115 also offers a low-cost RAID 0/1/10.
The IBM ServeRAID M1115 adapter has the following specifications:
򐂰 PCI Low Profile, half-length - MD2 form factor 򐂰 A total of 8 internal 6 Gbps SAS/SATA ports 򐂰 6 Gbps throughput per port 򐂰 533 MHz IBM PowerPC® processor with LSI SAS2008 6 Gbps RAID on Chip
(ROC) controller
򐂰 PCI Express 2.0 x8 host interface 򐂰 Support for RAID levels 0, 1, and 10 standard 򐂰 Supports RAID levels 5 and 50 with optional M1100 Series RAID 5 upgrade,
81Y4542
򐂰 Support for SAS and SATA HDDs and SSDs 򐂰 Support for intermixing SAS and SATA HDDs and SSDs 򐂰 Support for up to 16 virtual drives, up to 16 drive groups, up to 16 virtual
drives per one drive group, and up to 16 physical drives per one drive group
򐂰 Support for virtual drive sizes up to 64 TB 򐂰 Configurable stripe size up to 64 KB 򐂰 Compliant with Disk Data Format (DDF) configuration on disk (COD)
82 IBM NeXtScale System Planning and Implementation Guide
򐂰 S.M.A.R.T. support 򐂰 MegaRAID Storage Manager management software
N2115 SAS/SATA HBA
The N2115 SAS/SATA HBA for IBM System x is an ideal solution for System x servers that require high-speed internal storage connectivity. This eight-port host bus adapter supports direct attachment to SAS and SATA internal HDDs and SSDs. With a low-profile form-factor design, the N2115 SAS/SATA HBA offers two x4 internal mini-SAS connectors.
The N2115 SAS/SATA HBA has the following features and specifications:
򐂰 LSI SAS2308 6 Gbps I/O controller 򐂰 PCI low profile, half-length - MD2 form factor 򐂰 PCI Express 3.0 x8 host interface 򐂰 Eight internal 6 Gbps SAS/SATA ports (support for 6, 3, or 1.5 Gbps speeds) 򐂰 Up to 6 Gbps throughput per port 򐂰 Two internal x4 Mini-SAS connectors (SFF-8087) 򐂰 Non-RAID (JBOD mode) support for SAS and SATA HDDs and SSDs (RAID
not supported)
򐂰 Optimized for SSD performance 򐂰 High-performance IOPS LSI Fusion-MPT architecture 򐂰 Advanced power management support 򐂰 Support for SSP, SMP, STP, and SATA protocols 򐂰 End-to-End CRC with Advanced Error Reporting 򐂰 T-10 Protection Model for early detection of and recovery from data corruption 򐂰 Spread Spectrum Clocking for EMI reductions
For more information, see the list of IBM Redbooks Product Guides in the RAID adapters category at this website:
http://www.redbooks.ibm.com/portals/systemx?Open&page=pg&cat=raid
Loading...