IBM z-OS User Manual

IBM ^ zSeries 990 and z/OS Reference Guide

August 2004

Table of Contents

z/Architecture page 5

IBM ~ zSeries 990 page 6

z990 Family Models page 10

z990 and z900 Performance Comparison page 12

z990 I/O SubSystem page 13

z990 Channels and I/O Connectivity page 15

Fibre Channel Connectivity page 17

Open Systems Adapter-Express Features

(OSA-Express) page 21

HiperSockets page 26

Cryptography page 28

Availability page 29

Advanced Availability Functions page 31

Parallel Sysplex Cluster Technology page 32

z990 Support for Linux page 46

zSeries 990 Family Conﬁ guration Detail page 50

Physical Characteristics page 53

Coupling Facility - CF Level of Support page 54

z/OS page 56

z/VM page 79

To Learn More page 88

zSeries Overview

Technology has always accelerated the pace of change.

New technologies enable new ways of doing business,

shifting markets, changing customer expectations, and

redeﬁ ning business models. Each major enhancement to

technology presents opportunities. Companies that under-

stand and prepare for changes can gain advantage over

competitors and lead their industries.

Customers of every size, and in every industry are looking

for ways to make their businesses more resilient in the face

of change and uncertainty. They want the ability to react to

rapidly changing market conditions, manage risk, outpace

their competitors with new capabilities and deliver clear

returns on investments.

Welcome to the on demand era, the next phase of

e-business, in which companies move beyond simply

integrating their processes to actually being able to sense

and respond to ﬂ uctuating market conditions and provide

products and services to customers on demand. While the

former notion of on demand as a utility capability is a key

component, on demand companies have much broader

capabilities.

What does an on demand company look like?

• Responsive: It can sense and respond in real time to the changing needs of customers, employees, suppliers and partners

• Variable: It must be capable of employing variable cost structures to do business at high levels of productivity,

cost control, capital efﬁ ciency and ﬁ nancial predictability.

• Focused: It concentrates on its core competencies – areas where it has a differentiating advantage – and draws on the skills of strategic partners to manage needs outside of these competencies.

• Resilient: It can handle the ups and downs of the global market, and manage changes and threats with consistent availability, security and privacy – around the world, around the clock.

To support an on demand business, the IT infrastructure

must evolve to support it. At its heart the data center must

transition to reﬂ ect these needs; the data center must be

responsive to changing demands, it must be variable to

support the diverse environment, it must be ﬂ exible so that

applications can run on the optimal resources at any point

in time, and it must be resilient to support an always open

for business environment.

The on demand era plays to the strengths of the IBM

zSeries®. The IBM ^ zSeries 900

(z900) was launched in 2000 and was the ﬁ rst IBM server

"designed from the ground up for e-business." The latest

member of the family, the IBM ^ zSeries 990 (z990),

brings enriched functions that are required for the on

demand data center.

The "responsive" data center needs to have systems that

are managed to the quality of service goals of the business;

they need systems that can be upgraded transparently

to the user and they must be adaptable to the changing

requirements of the business. With the zSeries you have a

server with high levels of reliability and a balanced design

to ensure high levels of utilization and consistently high

service to the user. The capacity on demand features con-

tinue to evolve, helping to ensure that upgrading the servers

is timely and meets the needs of your business. It’s not

just the capacity of the servers that can be changed on

demand, but also the mix of workload and the allocation of

resources to reﬂ ect the evolving needs and priorities of the

business.

The variable data center needs to be able to respond to

the ever-changing demands that occur when you sup-

port multiple diverse workloads as a single entity. It must

respond to maintain the quality of service required and

the cost of utilizing the resources must reﬂ ect the chang-

environment. The zSeries Intelligent Resource Director

ing

(IRD), which combines three key zSeries technologies,

Workload Manager (WLM), Logical Partitioning and

z/OS

Parallel Sysplex

technology, helps ensure that your most

important workloads get the resources they need and con-

stantly manages the resources according to the changing

priorities of the business. With Workload License Charges

(WLC), as the resources required by different applications,

middleware and operating systems change over time,

the software costs change to reﬂ ect this. In addition, new

virtual Linux servers can be added in just minutes with

zSeries virtualization technology to respond rapidly

to huge increases in user activity.

The ﬂ exible data center must be adaptable to support

change and ease integration. This is achieved through

a combination of open and industry standards along

with the adaptability to direct resources where they are

required. The zSeries, along with other IBM servers, has

been investing in standards for years. Key is the support

™

for Linux, but let’s not forget Java

and XML and industry

standard technologies, such as FCP, Ethernet and SCSI.

Finally the on demand data center must be designed to be

resilient. The zSeries has been renowned for reliability and

availability. The zSeries platform will help protect against

both scheduled and unscheduled outages, and GDPS

enables protection from loss of complete sites.

The New zSeries from IBM – Impressive Investment -

Unprecedented Performance

IBM’s ongoing investment in zSeries technology has pro-

duced a re-invention of the zSeries server — the z990. The

z990 makes the mainframe platform more relevant to cur-

rent business success than ever before. Developed at an

investment in excess of $1 billion, the new z990 introduces

a host of new beneﬁ ts to meet today’s on demand business.

The major difference is the innovative book structure of the

z990. This new packaging of processors, memory and I/O

connections allows you to add incremental capacity to a

zSeries server as you need it. This makes the z990 a ﬂ ex-

ible and cost-effective zSeries server to date.

IBM’s investment in zSeries doesn’t stop here. To solidify

the commitment to zSeries, IBM introduces the “Mainframe

Charter” that provides a framework for future investment

and a statement of IBM’s dedication to deliver ongoing

value to zSeries customers in their transformation to on

demand business.

Tools for Managing e-business

The IBM ^ product line is backed by a compre-

hensive suite of offerings and resources that provide value

at every stage of IT implementation. These tools can help

customers test possible solutions, obtain ﬁ nancing, plan

and implement applications and middleware, manage

capacity and availability, improve performance and obtain

technical support across the entire infrastructure. The

result is an easier way to handle the complexities and

rapid growth of e-business. In addition, IBM Global Ser-

vices experts can help with business and IT consulting,

business transformation and total systems management

services, as well as customized e-business solutions.

z/Architecture

The zSeries is based on the z/Architecture™, which is

designed to reduce bottlenecks associated with the lack

of addressable memory and automatically directs resources

to priority work through Intelligent Resource Director. The

z/Architecture is a 64-bit superset of ESA/390.

z/Architecture is implemented on the z990 to allow full

64-bit real and virtual storage support. A maximum 256

GB of real storage is available on z990 servers. z990 can

deﬁ ne any LPAR as having 31-bit or 64-bit addressability.

z/Architecture has:

• 64-bit general registers.

• New 64-bit integer instructions. Most ESA/390 architecture instructions with 32-bit operands have new 64-bit and 32- to 64-bit analogs.

• 64-bit addressing is supported for both operands and instructions for both real addressing and virtual addressing.

• 64-bit address generation. z/Architecture provides 64-bit virtual addressing in an address space, and 64-bit real addressing.

• 64-bit control registers. z/Architecture control registers can specify regions, segments, or can force virtual addresses to be treated as real addresses.

• The preﬁ x area is expanded from 4K to 8K bytes.

New instructions provide quad-word storage consistency.

•

• The 64-bit I/O architecture allows CCW indirect data addressing to designate data addresses above 2 GB for both format-0 and format-1 CCWs.

• IEEE Floating Point architecture adds twelve new instructions for 64-bit integer conversion.

• The 64-bit SIE architecture allows a z/Architecture server to support both ESA/390 (31-bit) and z/Architecture (64-bit) guests. Zone Relocation is expanded to 64-bit

for LPAR and z/VM

• 64-bit operands and general registers are used for all Cryptographic instructions

• The implementation of 64-bit z/Architecture can help reduce problems associated with lack of addressable memory by making the addressing capability virtually unlimited (16 Exabytes).

z/Architecture Operating System Support

The z/Architecture is a tri-modal architecture capable of

executing in 24-bit, 31-bit, or 64-bit addressing modes.

Operating systems and middleware products have been

modiﬁ ed to exploit the new capabilities of the z/Architecture

Immediate beneﬁ t can be realized by the elimination of the

overhead of Central Storage to Expanded Storage page

movement and the relief provided for those constrained by

the 2 GB real storage limit of ESA/390. Application programs

can run unmodiﬁ ed on the zSeries family of servers.

Expanded Storage (ES) is still supported for operating sys-

tems running in ESA/390 mode (31-bit). For z/Architecture

mode (64-bit), ES is supported by z/VM. ES is not supported

by z/OS in z/Architecture mode.

Although z/OS does not support Expanded Storage when

running under the new architecture, all of the Hiperspace

™

and VIO APIs, as well as the Move Page (MVPG) instruc-

tion, continue to operate in a compatible manner. There is

no need to change products that use Hiperspaces.

Some of the exploiters of z/Architecture for z/OS include:

• DB2 Universal Database™ Server for z/OS

™

• IMS

• Virtual Storage Access Method (VSAM)

• Remote Dual Copy (XRC)

• Tape and DASD access method

IBM

zSeries 990

Operating System ESA/390 z/Arch Compati Exploita (31-bit) (64-bit) bility tion

OS/390® Version 2 Release 10 Yes Yes Yes No

z/OS Version 1 Release 2 No* Yes Yes No

z/OS Version 1 Release 3 No* Yes Yes No

z/OS Version 1 Release 4 No* Yes Yes Yes

z/OS Version 1 Release 5, 6 No Yes Included Included

Linux on S/390

Linux on zSeries No Yes Yes Yes

z/VM Version 3 Release 1 Yes Yes Yes No

z/VM Version 4 Release 3 Yes Yes Yes No

z/VM Version 4 Release 4 Yes Yes Included Yes

z/VM Version 5 Release 1 (3Q04)

VSE/ESA

z/VSE Version 3 Release 1 Yes No Yes Yes

TPF Version 4 Release 1 Yes No Yes No

(ESA mode only)

* Customers with z/OS Bimodal Migration Accommodation Offering may run in 31-bit support per the terms and conditions of the Offering. Bimodal Offering available for z/OS ONLY.

Yes No Yes Yes

No Yes Included Yes

™

Ver. 2 Release 6, 7 Yes No Yes Yes

IBM ^ zSeries is the enterprise class e-business

server optimized for the integration, transactions and data

of the next generation e-business world. In implement-

ing the z/Architecture with new technology solutions, the

zSeries models are designed to facilitate the IT business

transformation and reduce the stress of business-to-busi-

ness and business-to-customer growth pressure. The

zSeries represents an advanced generation of servers

that feature enhanced performance, support for zSeries

Parallel

Sysplex

clustering, improved hardware man-

agement controls and innovative functions to address

e-business processing.

The z990 server

technology

enhances performance by exploiting new

through many design enhancements. With a

new superscalar microprocessor and the CMOS 9S-SOI

technology, the z990 is designed to further extend and

integrate key platform characteristics such as dynamic

ﬂ exible partitioning and resource management in mixed

and unpredictable

ability, high availability

e-business applications such as WebSphere

workload environments, providing scal-

and Quality of Service to emerging

, Java and

Linux.

The z990 has 4 models available as new build systems

and as upgrades from the z900.

The four z990 models are designed with a multi-book

system structure which provides up to 32 Processor Units

(PUs) that can be characterized prior to the shipment of the

machine as either Central Processors (CPs), Integrated

Facility for Linux (IFLs), or Internal Coupling Facilities (ICFs).

The new IBM ^ zSeries Application Assist

Processor (zAAP), planned to be available on the IBM

^ zSeries 990 (z990) and zSeries 890 (z890) serv-

ers, is an attractively priced specialized processing unit

that provides strategic z/OS Java execution environment

for customers who desire the powerful integration advan-

tages and traditional Qualities of Service a of the zSeries

platform.

tures and improve operational efﬁ ciencies. For example,

use of zAAPs to strategically integrate Java Web applica-

tions with backend databases could reduce the number of

TCP/IP programming stacks, ﬁ rewalls, and physical inter-

connections (and their associated processing) that might

otherwise be required when the application servers and

their database servers are deployed on separate physical

server platforms.

When conﬁ gured with general purpose Central Processors

(CPs) within logical partitions running z/OS, zAAPs can

help you to extend the value of your existing zSeries

investments and strategically integrate and run e-business

Java workloads on the same server as your database,

helping to simplify and reduce the infrastructure required

for Web applications while helping to lower your overall

total cost of ownership.

zAAPs are designed to operate asynchronously with the

general purpose CPs to execute Java programming under

control of the IBM Java Virtual Machine (JVM). This can

help reduce the demands and capacity requirements

on general purpose CPs which may then be available

for reallocation to other zSeries workloads. The amount

of general purpose CP savings may vary based on the

amount of Java application code executed by zAAP(s).

And best of all, IBM JVM processing cycles can be

executed on the conﬁ gured zAAPs with no anticipated

modiﬁ cations to the Java application(s). Execution of the

JVM processing cycles on a zAAP is a function of the IBM

Software Developer’s Kit (SDK) for z/OS Java 2 Technology

Edition, z/OS 1.6 (or z/OS.e 1.6) and the innovative

™

Processor Resource/Systems Manager

(PR/SM™).

Notably execution of the Java applications on zAAPs,

within the same z/OS LPAR as their associated database

subsystems, can also help simplify the server infrastruc-

Essentially, zAAPs allow customers to purchase additional

processing power exclusively for z/OS Java application

execution without affecting the total MSU rating or machine

model designation. Conceptually, zAAPs are very similar to

a System Assist Processor (SAP); they cannot execute an

Initial Program Load and only assist the general purpose

CPs for the execution of Java programming. Moreover,

IBM does not impose software charges on zAAP capacity.

Additional IBM software charges will apply when additional

general purpose CP capacity is used.

Customers are encouraged to contact their speciﬁ c ISVs/

USVs directly to determine if their charges will be affected.

With the introduction of the z990, customers can expect to

see the following performance improvements:

Number of CPs Base Ratio

1 z900 2C1 1.54 - 1.61

8 z900 2C8 1.52 - 1.56

16 z900 2C16 1.51 - 1.55

32 z900 2C16 2.46 - 2.98

Note: Greater than 16 CPs requires a minimum of two operating system images

The Large System Performance Reference (LSPR) should

be referenced when considering performance on the z990.

Visit: ibm.com/servers/eserver/zseries/lspr/ for more infor-

mation on LSPR.

To support the new scalability of the z990 a new improve-

ment to the I/O Subsystem has been introduced to “break

the barrier” of 256 channels per Central Electronic Com-

plex (CEC). This provides “horizontal” growth by allowing

the deﬁ nition of up to four Logical Channel SubSystems

each capable of supporting up to 256 channels giving a

total of up to 1024 CHPIDs per CEC. The increased scal-

ability is further supported by the increase in the number

of Logical Partitions available from the current 15 LPARs to

a new 30 LPARs. There is still a 256-channel limit per oper-

ating system image.

These are some of the signiﬁ cant enhancements in the

zSeries 990 server that bring improved performance, avail-

ability and function to the platform. The following sections

highlight the functions and features of the server.

z990 Design and Technology

The z990 is designed to provide balanced system perfor-

mance. From processor storage to the system’s I/O and

network channels, end-to-end bandwidth is provided and

designed to deliver data where and when it is needed.

The z990 provides a signiﬁ cant increase in system scal-

ability and opportunity for server consolidation by pro-

viding four models, from one to four MultiChip Modules

(MCMs), delivering up to a maximum 32-way conﬁ gura-

tion. The MCMs are conﬁ gured in a book package, with

each book comprised of a MultiChip Module (MCM),

memory cards and Self-Timed Interconnects. The MCM,

which measures approximately 93 x 93 millimeters (42%

smaller than the z900), contains the processor unit (PU)

chips, the cache structure chips and the processor stor-

age controller chips. The MCM contains 101 glass ceramic

layers to provide interconnection between the chips and

the off-module environment. In total, there is approximately

0.4 kilometer of internal copper wiring on this module.

This new MCM packaging delivers an MCM 42% smaller

than the z900, with 23% more I/O connections and 133%

I/O density improvement. Each MCM provides support for

12 PUs and 32 MB level 2 cache. Each PU contains 122

million transistors and measures 14.1 mm x 18.9 mm. The

design of the MCM technology on the z990 provides the

ﬂ exibility to conﬁ gure the PUs for different uses; two of

the PUs are reserved for use as System Assist Processors

(SAPs), two are reserved as spares. The remaining inac-

tive 8 PUs on the MCM are available to be characterized

as either CPs, ICF processors for Coupling Facility appli-

cations, IFLs for Linux applications, IBM ^ zSeries

Application Assist Processor (zAAPs) for Java applications

or as optional SAPs, providing the customer with tremen-

dous ﬂ exibility in establishing the best system for running

applications. Each model of the z990 must always be

ordered with at least one CP, IFL or ICF.

The PU, which uses the latest chip technology from IBM

semiconductor laboratories, is built on CMOS 9S-SOI with

copper interconnections. The 14.1 mm x 18.9 mm chip has

a cycle time of 0.83 nanoseconds. Implemented on this

chip is the z/Architecture with its 64-bit capabilities includ-

ing instructions, 64-bit General Purpose Registers and

translation facilities.

Each book can support up to 64 GB of Memory, delivered

on two memory cards, and 12 STIs giving a total of 256 GB

of memory and 48 STIs on the D32 model. The memory is

delivered on 8 GB, 16 GB or 32 GB memory cards which

can be purchased in 8 GB increments. The minimum

memory is 16 GB. The two memory cards associated with

each book must be the same size. Each book has 3 MBAs

and each MBA supports 4 STIs.

All books are interconnected with a super-fast bi-direc-

tional redundant ring structure which allows the system to

be operated and controlled by PR/SM operating in LPAR

mode as a symmetrical, memory coherent, multiproces-

sor. PR/SM provides the ability to conﬁ gure and operate

as many as 30 Logical Partitions which may be assigned

processors, memory and I/O resources from any of the

available books. The z990 supports LPAR mode only (i.e.

basic mode is no longer supported).

The MultiChip Module is the technology cornerstone for

ﬂ exible PU deployment in the z990 models. For most

models, the ability of the MCM to have inactive PUs allows

such features as Capacity Upgrade on Demand (CUoD),

Customer Initiated Upgrades (CIU), and the ability to add

CPs, ICFs, IFLs, and zAAPs dynamically providing nondis-

ruptive upgrade of processing capability. Also, the ability

to add CPs lets a z990 with spare PU capacity become a

backup for other systems in the enterprise; expanding the

z990 system to meet an emergency outage situation. This

is called Capacity BackUp (CBU). The greater capacity of

the z990 offers customers even more ﬂ exibility for using

this feature to backup critical systems in their enterprise.

In order to support the highly scalable multi-book system

design the I/O SubSystem has been enhanced by intro-

ducing a new Logical Channel SubSystem (LCSS) which

provides the capability to install up to 1024 CHPIDs across

three I/O cages (256 per operating system image). I/O

improvements in the Parallel Sysplex Coupling Link archi-

tecture and technology support faster and more efﬁ cient

transmission between the Coupling Facility and production

™

systems. HiperSockets

provides high-speed capability to

communicate among virtual servers and Logical Partitions;

this is based on high-speed TCP/IP memory speed trans-

fers and provides value in allowing applications running

in one partition to communicate with applications running

in another without dependency on an external network.

Industry standard and openness are design objectives for

I/O in z990. The improved I/O subsystem is delivering new

horizons in I/O capability and has eliminated the 256 limit

to I/O attachments for a mainframe.

z990 Family Models

The z990 offers 4 models, the A08, B16, C24 and D32,

which can be conﬁ gured to give customers a highly scal-

able solution to meet the needs of both high transaction

processing applications and the demands of e-business.

The new model structure provides between 1-32 conﬁ gu-

rable Processor Units (PUs) which can be characterized

as either CPs, IFLs, ICFs, or zAAPs. A new easy-to-enable

ability to “turn off” CPs is available on z990 (a similar offer-

ing was available via RPQ on z900). The objective is to

allow customers to purchase capacity for future use with

minimal or no impact on software billing. An MES feature

will enable the CPs for use where the customer requires

the increased capacity. There are a wide range of upgrade

options available which are indicated in the z990 models

chart.

Unlike other zSeries server offerings, it is no longer pos-

sible to tell by the hardware model (A08, B16, C24, D32)

the number of PUs that are being used as CPs. For soft-

ware billing purposes only, there will be a “software” model

associated with the number of PUs that are characterized

as CPs. This number will be reported by the Store System

Information (STSI) instruction for software billing purposes

only. There is no afﬁ nity between the hardware model and

the number of CPs. For example, it is possible to have a

Model B16 which has 5 PUs characterized as CPs, so for

software billing purposes, the STSI instruction would report

305. The more normal conﬁ guration for a 5-way would be

an A08 with 5 PUs characterized as CPs. The STSI instruc-

tion would also report 305 for that conﬁ guration.

z990 Models

*S/W

301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332

Model

Model A08

Model B16

Model C24

* S/W Model refers to number of installed CPs. Reported by STSI instruction. Model 300 does not have any CPs. Note: For MSU values, refer to: ibm.com/servers/eserver/zseries/library/swpriceinfo/

Model D32

z990 and IBM ^ On/Off Capacity on Demand

IBM ^ On/Off Capacity on Demand (On/Off CoD)

is offered with z990 processors to provide a temporary

increase in capacity to meet customer's peak workload

requirements. The scope of On/Off Capacity on Demand

is to allow customers to temporarily turn on unassigned/

unowned PUs available within the current model for use as

CPs or IFLs. Temporary use of CFs, memory and channels

is not supported.

Before customers can order temporary capacity, they must

have a signed agreement for Customer Initiated upgrade

(CIU) facility. In addition to that agreement, they must

agree to speciﬁ c terms and conditions which govern the

use of temporary capacity.

Typically, On/Off Capacity on Demand will be ordered

through CIU, however there will be an RPQ available if no

RSF connection is present.

Although CBU and On/Off Capacity on Demand can both

reside on the server, the activation of On/Off Capacity

on Demand is mutually exclusive with Capacity BackUp

(CBU) and no physical hardware upgrade will be sup-

ported while On/Off Capacity on Demand is active.

This important new function for zSeries gives customers

greater control and ability to add capacity to meet the

requirements of an unpredictable on demand applica-

tion environment. On/Off CoD extends zSeries capacity

on demand offerings to the next level of ﬂ exibility. It is

designed to help customers match cost with capacity

utilization and manage periodic business spikes. On/Off

Capacity on Demand is designed to provide a low-risk way

to deploy pilot applications, and it is designed to enable a

customer to grow capacity rationally and proportionately

with market demand.

and up to 360 ESCON channels. Each book will support

up to 12 STIs for I/O connectivity. Seven STIs are required

to support the 28 channel slots in each I/O cage so in

order to support a fully conﬁ gured three I/O cage system

21 STIs are required. To achieve this maximum I/O con-

nectivity requires at least a B16 model which provides 24

STIs.

The following chart shows the upgrade from z900 to z990.

There are any to any upgrades from any of the z900 gen-

eral purpose models. A z900 Coupling Facility Model 100

must ﬁ rst be upgraded to a z900 general purpose model

before upgrading to a z990. There are no upgrades from

9672 G5/G6 or IBM ^ zSeries 800 (z800).

Model Upgrades

z900 z990

Customers can also take advantage of Capacity Upgrade

on Demand (CUoD), Customer Initiated Upgrade (CIU),

and Capacity BackUp (CBU) which are described later in

the document.

The z990 has also been designed to offer a high perfor-

mance and efﬁ cient I/O structure. All z990 models ship

with two frames the A-Frame and the Z-Frame; this sup-

ports the installation of up to three I/O cages. Each I/O

cage has the capability of plugging up to 28 I/O cards.

When used in conjunction with the software that supports

Logical Channel SubSystems, it is possible to have up to

420 ESCON

mum of 1024 channels across 3 I/O cages. Alternatively,

three I/O cages will support up to 120 FICON

channels in a single I/O cage and a maxi-

™

channels

100

101 - 109 A08

1C1 - 116

2C1 - 216

B16

C24

D32

z990 and z900 Performance Comparison

The performance design of the z/Architecture enables the

entire server to support a new standard of performance for

applications through expanding upon a balanced system

approach. As CMOS technology has been enhanced to

support not only additional processing power, but also

more engines, the entire server is modiﬁ ed to support the

increase in processing power. The I/O subsystem supports

a great amount of bandwidth through internal changes,

thus providing for larger and quicker data movement into

and out of the server. Support of larger amounts of data

within the server required improved management of stor-

age conﬁ gurations made available through integration of

the software operating system and hardware support of

64-bit addressing. The combined balanced system effect

allows for increases in performance across a broad spec-

trum of work. However, due to the increased ﬂ exibility in

the z990 model structure and resource management in

the system, it is expected that there will be larger perfor-

mance variability than has been previously seen by our

traditional customer set. This variability may be observed

in several ways. The range of performance ratings across

the individual LSPR workloads is likely to have a larger

spread than past processors. There will also be more

performance variation of individual LPAR partitions as the

impact of ﬂ uctuating resource requirements of other parti-

tions can be more pronounced with the increased number

of partitions and additional CPs available on the z990. The

customer impact of this increased variability will be seen

as increased deviations of workloads from single-number-

metric based factors such as MIPS, MSUs and CPU time

chargeback algorithms. It is important to realize the z990

has been optimized to run many workloads at high utiliza-

tion rates.

It is also important to notice that the LSPR workloads for

z990 have been updated to reﬂ ect more closely our cus-

tomers’ current and growth workloads. The traditional TSO

LSPR workload is replaced by a new, heavy Java tech-

nology-based online workload referred to as Trade2-EJB

(a stock trading application). The traditional CICS

/DB2®

LSPR online workload has been updated to have a Web-

frontend which then connects to CICS. This updated

workload is referred to as WEB/CICS/DB2 and is repre-

sentative of customers who Web-enable access to their

legacy applications. Continuing in the LSPR for z990 will

be the legacy online workload, IMS, and two legacy batch

workloads CB84 and CBW2. The z990 LSPR will provide

performance ratios for individual workloads as well as a

“default mixed workload” which is used to establish single-

number-metrics such as MIPS, MSUs and SRM constants.

The z990 default mixed workload will be composed of

equal amounts of ﬁ ve workloads, Trade2-EJB, WEB/CICS/

DB2, IMS, CB84 and CBW2. Additionally, the z990 LSPR

will rate all z/Architecture processors running in LPAR

mode and 64-bit mode. The existing z900 processors have

all been re-measured using the new workloads — all run-

ning in LPAR mode and 64-bit mode.

Using the new LSPR ‘default mixed workload’, and with all

processors executing in 64-bit and LPAR mode, the follow-

ing results have been estimated:

• Comparing a one-way z900 Model 2C1 to a z990 model with one CP enabled, it is estimated that the z990 model has 1.52 to 1.58 times the capacity of the 2C1.

• Comparing an 8-way z900 Model 2C8 to a z990 model with eight CPs enabled, it is estimated that the z990 model has 1.48 to 1.55 times the capacity of the 2C8.

z990 I/O SubSystem

• Comparing a 16-way z900 Model 216 to a z990 model with sixteen CPs enabled, it is estimated that the z990 model has 1.45 to 1.53 times the capacity of the 216.

• Comparing a 16-way z900 Model 216 to a z990 model with thirty-two CPs enabled, and the workload executing on the z990 executing in two 16-way LPARs, it is estimated that the z990 model has 2.4 to 2.9 times the capacity of the 216.

Model D32

Model C24

Model B16

Model A08

z900

2C2 2C3 2C4 2C5 2C6 2C7 2C8 2C9 210 211 212 213 214 215 2162C1

Turbo

1C2 1C3 1C4 1C5 1C6 1C7 1C8 1C9 110 111 112 113 114 115 1161C1

z900

Note: Expected performance improvements are based on hardware changes. Additional performance beneﬁ ts may be obtained as the z/Architecture is fully exploited.

The z990 contains an I/O subsystem infrastructure which

uses an I/O cage that provides 28 I/O slots and the abil-

ity to have one to three I/O cages delivering a total of 84

™

I/O slots. ESCON, FICON Express

and OSA-Express

features plug into the z990 I/O cage along with any ISC-

3s, STI-2 and STI-3 distribution cards, and PCICA and

PCIXCC features. All I/O features and their support cards

can be hot-plugged in the I/O cage. Installation of an I/O

cage remains a disruptive MES, so the Plan Ahead feature

remains an important consideration when ordering a z990

system. The A08 model has 12 available STIs and so has

connectivity to a maximum of 12 I/O domains, i.e. 48 I/O

slots, so if more than 48 I/O slots are required a Model B16

is required. Each model ships with one I/O cage as stan-

dard in the A-Frame (the A-Frame also contains the pro-

cessor CEC cage) any additional I/O cages are installed in

the Z-Frame. The z990 provides a 400 percent increase in

I/O bandwidth provided by the STIs.

z990 Cage Layout

3rd

I/O Cage

2nd

I/O Cage

A-FrameZ-Frame

CEC

1st

I/O Cage

z990 Logical Channel SubSystems (LCSSs) and support for

greater than 15 Logical Partitions (LP)

In order to provide the increased channel connectivity

required to support the scalability of the z990, the z990

channel I/O SubSystem delivers a breakthrough in con-

nectivity, by providing up to 4 LCSS per CEC, each of

which can support up to 256 CHPIDs with exploitation soft-

ware installed. This support is provided in such a way that

is transparent to the programs operating in the logical par-

tition. Each Logical Channel SubSystem may have from 1

to 256 CHPIDs and may in turn be conﬁ gured with 1 to 15

logical partitions. Each Logical Partition (LPAR) runs under

a single LCSS. As with previous zSeries servers, Multiple

Image Facility (MIF) channel sharing as well as all other

channel subsystem features are available to each Logical

Partition conﬁ gured to each Logical Channel SubSystem.

Physical Channel IDs (PCHIDs) SubSystem

In order to accommodate the new support for up to 1024

CHPIDs introduced with the Logical Channel SubSystem

(LCSS) a new Physical Channel ID (PCHID) is being intro-

duced. The PCHID represents the physical location of an

I/O feature in the I/O cage. CHPID numbers are no longer

pre-assigned and it is now a customer responsibility to do

this assignment via IOCP/HCD. CHPID assignment is done

by associating a CHPID number with a physical location,

the PCHID. It is important to note that although it is pos-

sible to have LCSSs, there is still a single IOCDS to deﬁ ne

the I/O subsystem. There is a new CHPID mapping tool

available to aid in the mapping of CHPIDs to PCHIDs. The

™

CHPID Mapping tool is available from Resource Link

, at

ibm.com/servers/resourcelink.

Up to 256

LCSS0 LCSS1

CHPIDs

Up tp 30 Logical Partitions

Up to 256

CHPIDs

LCSS2

Up to 256

CHPIDs

LCSS3

Up to 256

CHPIDs

IOCP - IOCDS

Partitions Partitions

LCSS0

CHPIDs CHPIDs

12 52 EF4F 12 2F EF00 02

LCSS1

HCD - HSA or IOCDS - HSA

102 103 104 110 200 201 202 2B0 2C5

Physical Channels (PCHIDs)

Note: Crypto no longer requires a CHPID

z990 Channels and I/O Connectivity

Logical Channel SubSystem (LCSS) Spanning

The concept of spanning channels provides the ability for

a channel to be conﬁ gured to multiple LCSSs and therefore

they may be transparently shared by any/all of the logical

partitions in those LCSSs. Normal Multiple Image Facility

(MIF) sharing of a channel is conﬁ ned to a single LCSS.

The z990 supports the spanning of the channels types: IC,

HiperSockets, FICON Express, OSA-Express. Note: ESCON

architecture helps prevent the spanning of ESCON channels.

IC Channel Spanning

PRD1 PRD2 PRD3 PRD4 LNX1

ICF1

LCSS0 LCSS1

CHPID 22

A z990 with all three I/O cages installed has a total of 84

I/O slots. These slots can be plugged with a mixture of fea-

tures providing the I/O connectivity, networking connectiv-

ity, coupling and cryptographic capability of the server.

Up to 1024 ESCON Channels

The high density ESCON feature has 16 ports, 15 of which

can be activated for customer use. One port is always

reserved as a spare which is activated in the event of a

failure of one of the other ports. For high availability the ini-

tial order of ESCON features will deliver two cards and the

active ports will be distributed across those cards. After

the initial install the ESCON features are installed in incre-

ments of one. ESCON channels are available in four-port

increments and are activated using IBM Licensed Internal

Code, Conﬁ guration Control (LIC CC).

Up to 120 FICON Express Channels

The z990 supports up to 120 FICON Express channels.

FICON Express is available in long wavelength (LX) and

short wavelength (SX) features. Each FICON Express fea-

ture has two independent ports which support two chan-

nels per card. The LX and SX cannot be intermixed on a

single feature. The maximum number of FICON Express

features is 60 which can be installed across three I/O

cages with a maximum of 20 features per I/O cage.

The z990 supports up to 120 FCP channels for attach-

ment to Small Computer System Interface (SCSI) disks in

a Linux environment. The same two-port FICON Express

feature card used for FICON Express channels can also

be used for Fibre Channel Protocol (FCP) channels. FCP

channels are enabled on these existing features via a

License Internal Code (LIC) with a new mode of operation

and new CHPID deﬁ nition. FCP is available in long wave-

length (LX) and short wavelength (SX) features, though

the LX and SX cannot be intermixed on a single feature.

Note, the maximum quantity of FICON Express, OSA-

Express, PCICA, and PCIXCC features in combination

cannot exceed 20 features per I/O cage and 60 features

per server.

InterSystem Channel-3 (ISC-3)

A four feature ISC-3 link is provided on the z990 family

of servers. It consists of a mother card with two daugh-

ter cards which have two links each. Each of the four is

capable of operation at 1 gigabit per second (Gbps) in

Compatibility Mode or 2 Gbps in Peer Mode up to an

unrepeated distance of 10 km (6.2 miles). The mode is

selected for each port via the CHPID type in the IOCDS.

The ports are orderable in one-port increments.

An RPQ card (8P2197) is available to allow ISC-3 dis-

tances up to 20 km. This card runs in Peer Mode at 1

Gbps and/or Compatibility Mode at 1 Gbps.

Integrated Cluster Bus-3 (ICB-3)

The ICB-3 feature is a coupling link used to provide high-

speed communication between a z990 Server and a z900

General Purpose Server or Model 100 Coupling Facility

over a short distance (less than 7 meters). The ICB-3 is

supported via an STI-3 card which resides in the I/O cage

and converts the 2.0 GBps input into two 1 GBps ICB-3s.

Integrated Cluster Bus-4 (ICB-4)

The ICB-4 feature is a coupling link used to provide high-

speed communication between a z990 and/or z890 server

over a short distance (less than 7 meters). The ICB-4 con-

sists of a link that attaches directly to a 2.0 GBps STI port

on the server and does not require connectivity to an I/O

cage.

Internal Coupling Channel (IC)

IC links emulate the coupling links between images within

a single server. IC links are deﬁ ned in the IOCP. There is no

physical channel involved. A z/OS image can connect to a

coupling facility on the same server using IC capabilities.

Integrated Cluster Bus-2 (ICB-2)

The ICB-2 feature is a coupling link used to provide high-

speed communication between a 9672 G5/G6 server and

a z990 server over a short distance (less than 7 meters).

The ICB-2 is supported via an STI-2 card which resides in

the I/O cage and converts the 2.0 GigaBytes per second

(GBps) input into two 333 MegaBytes per second (MBps)

ICB-2s. ICB-2 is not supported between z990 and other

zSeries servers. ICB-2s cannot be used to connect to a

z900 server.

Fibre Channel Connectivity

The on demand operating environment requires fast data

access, continuous data availability, and improved ﬂ ex-

ibility all with lower cost of ownership. The new increased

number of FICON Express features available on the z990

helps distinguish this new server family, further setting it

apart as enterprise class in terms of the number of simulta-

neous I/O connections available for these FICON Express

features.

FICON Express Channel Card Features

Performance

With its 2 Gigabit per second link data rate capability, the

FICON Express channel card feature (feature codes 2319,

2320) is the latest zSeries implementation for the Fibre

Channel Architecture. The FICON Express card has two

links and can achieve improved performance over the

previous generation FICON channel card. For example,

attached to a 100 MBps link (1 Gbps), a single FICON

Express feature conﬁ gured as a native FICON channel

is capable of supporting up to 7,200 I/O operations/sec

(channel is 100% utilized) and an aggregate total through-

put of 120 MBps on z990.

With 2 Gbps links, customers may expect up to 170 MBps

of total throughput. The 2 Gbps link data rates are appli-

cable to native FICON and FCP channels on zSeries only

and for full beneﬁ t, require 2 Gbps capable devices as

well. Customers can leverage this additional bandwidth

capacity to consolidate channels and reduce conﬁ guration

complexity, infrastructure costs, and the number of chan-

nels that must be managed. Please note, no additional

hardware or code is needed in order to obtain 2 Gbps

links. The functionality was incorporated in all zSeries

with March 2002 LIC. The link data rate is auto-negotiated

between server and devices.

Flexibility - Three channel types supported

The FICON Express features support three different chan-

nel types: 1) FCV Mode for FICON Bridge Channels, 2) FC

mode for Native FICON channels (including the FICON

CTC function), and 3) FCP mode for Fibre Channels (FCP

channels). Support for FCP devices means that zSeries

servers will be capable of attaching to select ﬁ bre channel

switches/directors and FCP/SCSI disks and may access

these devices from Linux on zSeries and, new with z/VM

Version 5 Release 1, installation and operation of z/VM on

a SCSI disk.

Distance

All channels deﬁ ned on FICON Express LX channel card

features at 1 Gbps link data rates support a maximum

unrepeated distance of up to 10 km (6.2 miles, or up to

20 km via RPQ, or up to 100 km with repeaters) over nine

micron single mode ﬁ ber and up to 550 meters (1,804

feet) over 50 or 62.5 micron multimode ﬁ ber through Mode

Conditioning Patch (MCP) cables. At 2 Gbps link speeds

FICON Express LX channel card features support up to

10 km (6.2 miles, or up to 12 km via RPQ, or up to 100

km with repeaters) over nine micron single mode ﬁ ber.

At 2 Gbps link speeds, Mode Conditioning Patch (MCP)

cables on 50 or 62.5 micron multimode ﬁ ber are not sup-

ported. The maximum unrepeated distances for 1 Gbps

links deﬁ ned on the FICON Express SX channel cards are

up to 500 meters (1,640 feet) and 250 meters (820 feet)

for 50 and 62.5 micron multimode ﬁ ber, respectively. The

maximum unrepeated distances for 2 Gbps links deﬁ ned

on the FICON Express SX channel cards are up to 300

meters and 120 meters for 50 and 62.5 micron multimode

ﬁ ber, respectively. The FICON Express channel cards are

designed to reduce the data droop effect that made long

distances not viable for ESCON. This distance capability is

becoming increasingly important as enterprises are moving

toward remote I/O, vaulting for disaster recovery and

™

Geographically Dispersed Parallel Sysplex

for availability.

Shared infrastructure

FICON (FC-SB-2 Fibre Channel Single-Byte Command

Code Set-2) has been adopted by INCITS (International

Committee for Information Technology Standards) as a

standard to the Fibre Channel Architecture. Using open

connectivity standards leads to shared I/O ﬁ ber cabling

and switch infrastructures, facilitated data sharing, storage

management and SAN implementation, and integration

between the mainframe and UNIX

/Intel® technologies.

Native FICON Channels

Native FICON channels and devices can help to reduce

bandwidth constraints and channel contention to enable

easier server consolidation, new application growth, large

business intelligence queries and exploitation of e-business.

Currently, the IBM TotalStorage

Enterprise Storage Server®

(ESS) Models F10, F20 and 800 have two host adapters

to support native FICON. These host adapters each have

one port per card and can either be FC 3021 for long

wavelength or FC 3032 for short wavelength on the F10/

F20, or FC 3024 for long wavelength and 3025 for short

wavelength on the 800. All three models can support up

to 16 FICON ports per ESS. The Model 800 is 2 Gb link

capable. The IBM TotalStorage Enterprise Tape Control-

3590 Model A60 provides up to two FICON interfaces

ler

which can coexist with ESCON on the same box. Enter-

prise Tape Controller 3592-J70 provides up to four FICON

interfaces, which can exist with ESCON on the same box.

The 3592-J70 is designed to provide up to 1.5 times the

throughput of the Model A60. Customers can utilize IBM’s

highest capacity, highest performance tape drive to sup-

port their new business models.

Many Fibre Channel directors provide dynamic connectiv-

ity to native FICON control units. The IBM 2032 models

001, 064 and 140 (resell of the McDATA ED-5000, and

Intrepid 6000 Series Directors) are 32-, 64- and 140-port

high availability directors. The IBM 2042 Models 001, 128

and 256 (resell of the CNT FC/9000 Directors) are 64-,

128- and 256-port high availability directors. All have fea-

tures that provide interface support to allow the unit to be

managed by System Automation for OS/390. The McDATA

Intrepid 6000 Series Directors and CNT FC/9000 Directors

support 2 Gbps link data rates as well.

The FICON Express features now support attachment to

the IBM M12 Director (2109-M12). The IBM M12 Director

supports attachment of FICON Express channels on the

z990 via native FICON (FC CHPID type) and Fibre Channel

Protocol (FCP CHPID type) supporting attachment to SCSI

disks in Linux environments.

Wave Division Multiplexor and Optical Ampliﬁ ers that sup-

port 2 Gbps FICON Express links are: Cisco Systems ONS

15530 and 15540 ESP (LX, SX) and optical ampliﬁ er (LX,

SX), Nortel Networks Optera Metro 5100, 5200 and 5300E

and optical ampliﬁ er, ADVA Fiber Service Platform (FSP)

2000 system and the IBM 2029 Fiber Saver.

The raw bandwidth and distance capabilities that native

FICON end-to-end connectivity has to offer makes them of

interest for anyone with a need for high performance, large

data transfers or enhanced multi-site solutions.

FICON Connectivity

FICON Bridge

ESCD

9032

Model 5

2032

FICON Bridge

32, 64 or 140

PORT

ESCON

ESS

F10, F20, 800

Enterprise

Tape

Controller

3590 A60

FICON CTC function

Native FICON channels support channel-to-channel (CTC)

on the z990, z890, z900 and z800. G5 and G6 servers

can connect to a zSeries FICON CTC as well. This FICON

CTC connectivity will increase bandwidth between G5, G6,

z990, z890, z900, and z800 systems.

Because the FICON CTC function is included as part of the

native FICON (FC) mode of operation on zSeries, FICON

CTC is not limited to intersystem connectivity (as is the case

with ESCON), but will also support multiple device deﬁ ni-

tions. For example, ESCON channels that are dedicated as

CTC cannot communicate with any other device, whereas

native FICON (FC) channels are not dedicated to CTC only.

Native can support both device and CTC mode deﬁ nition

concurrently, allowing for greater connectivity ﬂ exibility.

64, 128 or 256

All FICON Channels =

100MB/s

= LX ONLY

= LX or SX

2042

PORT

ESS

F10, F20, 800

Enterprise

Tape

Controller

3590 A60

FICON Support for Cascaded Directors

Native FICON (FC) channels now support cascaded

directors. This support is for a single hop conﬁ guration

only. Two-director cascading requires a single vendor

high integrity fabric. Directors must be from the same

vendor since cascaded architecture implementations can

be unique. This type of cascaded support is important

for disaster recovery and business continuity solutions

because it can help provide high availability, extended

distance connectivity, and (particularly with the implemen-

tation of 2 Gbps Inter Switch Links), has the potential for

ﬁ ber infrastructure cost savings by reducing the number of

channels for interconnecting the 2 sites.

FICON cascaded directors have the added value of high

integrity connectivity. New integrity features introduced

within the FICON Express channel and the FICON cas-

caded switch fabric to aid in the detection and reporting

of any miscabling actions occurring within the fabric can

prevent data from being delivered to the wrong end point.

FICON cascaded directors are offered in conjunction with

IBM, CNT, and McDATA directors.

FCP Channels

zSeries supports FCP channels, switches and FCP/SCSI

disks with full fabric connectivity under Linux on zSeries

and z/VM Version 4 Release 3 and later for Linux as a

guest under z/VM. Support for FCP devices means that

zSeries servers will be capable of attaching to select FCP/

SCSI devices and may access these devices from Linux

on zSeries. This expanded attachability means that enter-

prises have more choices for new storage solutions, or

may have the ability to use existing storage devices, thus

leveraging existing investments and lowering total cost of

ownership for their Linux implementation.

IBM

Two site non-cascaded director

topology. Each CEC connects to

directors in both sites.

With

Inter Switch Links (ISLs),

less fiber cabling may be needed

for cross-site connectivity

Two Site cascaded director

topology. Each CEC connects to

IBM

local directors only.

FICON Bridge Channel

Introduced ﬁ rst on the 9672 G5 processors, the FICON

Bridge (FCV) channel is still an effective way to use FICON

bandwidth with existing ESCON control units. FICON

Express LX channel cards in FCV (FICON Converted)

Mode of operation can attach to the 9032 Model 005

ESCON Director through the use of a director bridge card.

Up to 16 bridge cards are supportable on a single 9032

Model 005 with each card capable of sustaining up to

eight concurrent ESCON data transfers. 9032 Model 005

ESCON Directors can be ﬁ eld upgradeable at no charge

to support the bridge cards, and bridge cards and ESCON

cards can coexist in the same director.

For details of supported FICON and FCP attachments

access Resource Link at ibm.com/servers/resourcelink

and in the Planning section, go to z890/z990 I/O Connec-

tion information.

The support for FCP channels is for Linux on zSeries and

z/VM 4.3 and later for Linux as a guest under z/VM. Linux

may be the native operating system on the zSeries server

(note z990 runs LPAR mode only), or it can be in LPAR

mode or, operating as a guest under z/VM 4.3 or later. The

z990 now provides support for IPL of Linux guest images

from appropriate FCP attached devices.

Now, z/VM V5.1 supports SCSI FCP disks enabling the

deployment of a Linux server farm running under VM con-

ﬁ gured only with SCSI disks. With this support you can

install, IPL, and operate z/VM from SCSI disks.

The 2 Gbps capability on the FICON Express channel

cards means that 2 Gbps link speeds are available for

FCP channels as well.

Open Systems Adapter-Express Features (OSA-Express)

FCP Full fabric connectivity

FCP full fabric support means that any number of (single

vendor) FCP directors/ switches can be placed between

the server and FCP/ SCSI device, thereby allowing many

“hops” through a storage network for I/O connectivity.

This support along with 2 Gbps link capability is being

delivered together with IBM switch vendors IBM, CNT, and

McDATA. FCP full fabric connectivity enables multiple FCP

switches/ directors on a fabric to share links and there-

fore provides improved utilization of inter-site connected

resources and infrastructure. Further savings may be real-

ized in the reduction of the number of ﬁ ber optic cabling

and director ports.

When conﬁ gured as FCP CHPID type, the z990 FICON

Express features support the industry standard interface

for Storage Area Network (SAN) management tools.

FCP

Device

FCP

Device

Fibre Channel

Directors

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

FCP

Device

With the introduction of the z990, its increased process-

ing capacity, and the availability of Logical Channel

SubSystems, the OSA-Express Adapter family of Local

Area Network (LAN) features is also expanding by offer-

ing a maximum of up to 24 features per server, versus the

maximum of up to 12 features per server on prior genera-

tions. This expands the z990 balanced solution to increase

throughput and responsiveness in an on demand operat-

ing environment. These features combined with z/OS, or

OS/390, z/VM, Linux on zSeries, TPF, and VSE/ESA can

deliver a balanced system solution to increase throughput

and decrease host interrupts to help satisfy your business

goals.

Each of the OSA-Express features offers two ports for con-

nectivity delivered in a single I/O slot, with up to a maxi-

mum of 48 ports (24 features) per z990. Each port uses

a single CHPID and can be separately conﬁ gured. For a

new z990 build, you can choose any combination of OSA-

Express features: the new OSA-Express Gigabit Ethernet

LX or SX, the new OSA-Express 1000BASE-T Ethernet or

OSA-Express Token-Ring. The prior OSA-Express Gigabit

LX and SX, the OSA-Express Fast Ethernet, and the OSA-

Express Token-Ring can be carried forward on an upgrade

from z900.

z990 OSA-Express 1000BASE-T Ethernet

The new OSA-Express 1000BASE-T Ethernet feature

replaces the current Fast Ethernet (10/100 Mbps) feature.

This new feature is capable of operating at 10,100 or 1000

Mbps (1 Gbps) using the same copper cabling infrastructure

as Fast Ethernet making transition to this higher speed Ether-

net feature a straightforward process. It is designed to sup-

port Auto-negotiation, QDIO and non-QDIO environments on

each port, allowing you to make the most of your TCP/IP and

SNA/APPN® and HPR environments at up to gigabit speeds.

OSA-Express Integrated Console Controller

The new Open Systems Adapter-Express Integrated Con-

sole Controller function (OSA-ICC), which is exclusive to

IBM and the IBM z890 and z990 servers since it is based

on the OSA-Express feature, supports the attachment of

non-SNA 3270 terminals for operator console applications.

Now, 3270 emulation for console session connections

is integrated in the zSeries and can help eliminate the

requirement for external console controllers (2074, 3174)

helping to reduce cost and complexity.

When this adapter is operating at gigabit Ethernet speed it

runs full duplex only. It also can support standard (1492 or

1500 byte) and jumbo (8992 byte) frames.

The new Checksum ofﬂ oad support on the 1000BASE-T Eth-

ernet feature when operating in QDIO mode at gigabit speed

is designed to ofﬂ oad z/OS 1.5 and Linux TCP/IP stack pro-

cessing of Checksum packet headers for TCP/IP and UDP

non-QDIO mode SNA Passthru TCP/IP Passthru HPDT MPC

10/100/1000 Mbps Ethernet (copper)

Switch/

Hub/

Router

Server

IP WAN Intranet

DLSw

Router

10/100 Mbps

Ethernet

Server

10/100 Mbps

Ethernet

DLSw Router

Remote Office

4/16 Mbps

Token-Ring

SNA DLSw TCP/IP Native SNA

IP Router

10/100/1000 Mbps

Ethernet (copper)

10/100/1000

Mbps

Ethernet

IP WAN Intranet

10/100 Mbps

Ethernet

Server

QDIO Mode - TCP/IP

IBM ^ pSeries®,

RS/6000

IBM ^ xSeries

Netfinity

Switch/

Hub/

Router

10/100 Mbps

Server

Ethernet

IP Router

Internet or

extranet

Remote Office

4/16 Mbps

Token-Ring

TCP/IP applications TN3270 browser access to SNA applications Enterprise Extender for SNA end points

The OSA-ICC uses one or both ports on an OSA-Express

1000BASE-T Ethernet feature with the appropriate

Licensed Internal Code (LIC). The OSA-ICC is enabled

using CHPID type OSC.

The OSA-ICC supports up to 120 client console sessions

either locally or remotely.

Support for this new function will be available for z/VM

Version 4 Release 4 and later, z/OS Version 1 Release3,

VSE/ESA Version 2 Release 6 onwards and TPF.

Queued Direct Input/Output (QDIO)

The OSA-Express Gigabit Ethernet, 1000BASE-T Ethernet

and Token-Ring features support QDIO (CHPID type OSD),

which is unique to IBM. QDIO was ﬁ rst introduced to the

world on the z900, in Communication Server for OS/390 2.7.

Queued Direct Input/Output (QDIO), a highly efﬁ cient data

transfer architecture, breaks the barriers associated with

the Channel Control Word (CCW) architecture increasing

data rates and reducing CPU cycle consumption. QDIO

allows an OSA-Express feature to directly communicate

with the server’s communications program through the use

of data queues in memory. QDIO helps eliminate the use

of channel program and channel control words (CCWs),

helping to reduce host interrupts and accelerate TCP/IP

packet transmission.

TCP/IP connectivity is increased with the capability to

allow up to a maximum of 160 IP stacks per OSA-Express

port and 480 devices. This support is applicable to all the

OSA-Express features available on the z990 and is pro-

vided through the Licensed Code (LIC).

Full Virtual Local Area Network (VLAN) support is available

on z990 in z/OS V1.5 Communications Server (CS) for the

OSA-Express 1000BASE-T Ethernet, Fast Ethernet and

Gigabit Ethernet features when conﬁ gured in QDIO mode.

Full VLAN is also available with z/OS V1.2 on z900 and

z800 using appropriate LIC upgrade on Fast Ethernet and

Gigabit Ethernet features. Full VLAN support in a Linux on

zSeries environment was delivered for QDIO mode in April

2002 for z800 and z900. z/VM V4.4, and later, also exploits

the VLAN technology offering one global VLAN ID for IPv4.

z/VM V5.1 provides the support for one global VLAN ID for

IPv6.

z990 OSA-Express Gigabit Ethernet

The new z990 OSA-Express Gigabit Ethernet LX and Giga-

bit Ethernet SX features replace the z900 Gigabit Ethernet

features for new build z990. The new OSA-Express GbE

features have a new connector type, LC Duplex, replacing

the current SC Duplex connectors used on the prior z900

Gigabit Ethernet features. The new Checksum ofﬂ oad sup-

port on these z990 features is designed to ofﬂ oad z/OS

V1.5 and Linux on zSeries TCP/IP stack processing of

Checksum packet headers for TCP/IP and UDP.

QDIO Mode - TCP/IP

pSeries, RS/6000

Gigabit Ethernet

Gigabit

Ethernet

Gigabit Ethernet

Switch /

Router

Server

TCP/IP applications TN3270 browser access to SNA appls. Enterprise Extender for SNA end points

(fiber or copper)

10/100 Mbps

Server

IP WAN Intranet

4/16 Mbps

Token-Ring

4/16/100 Mbps

Ethernet

IP Router

Token-Ring

Internet or

Remote Office

10/100 Mbps

Ethernet

Server

xSeries, Netfinity

Server

IP Router

extranet

NON-QDIO operational mode

The OSA-Express 1000BASE-T Ethernet, Fast Ethernet

and Token-Ring also support the non-QDIO mode of

operation (CHPID type OSE). The adapter can only be set

(via the CHPID type parameter) to one mode at a time.

The non-QDIO mode does not provide the beneﬁ ts of

QDIO. However, this support includes native SNA/APPN,

High Performance Routing, TCP/IP passthrough, and

HPDT MPC. The new OSA-Express 1000BASE-T Ethernet

provides support for TCP/IP and SNA/APPN/HPR up to 1

gigabit per second over the copper wiring infrastructure.

z990 OSA-Express Token-Ring

The same OSA-Express Token-Ring feature is supported

on z990 and z900. This Token-Ring supports a range of

speed including 4, 16 and 100 Mbps, and can operate in

both QDIO and non-QDIO modes.

Note: OSA-Express 155 ATM and OSA-2 FDDI are no

longer supported. If ATM or FDDI support are still required,

a multiprotocol switch or router with the appropriate net-

work interface for example, 1000BASE-T Ethernet, GbE LX

or GbE SX can be used to provide connectivity between

the LAN and the ATM network or FDDI LAN.

Server to User connections

A key strength of OSA-Express and associated Commu-

nications Server protocol support is the ability to accom-

modate the customer’s attachment requirements, spanning

combinations of TCP/IP and SNA applications and

devices. Customers can use TCP/IP connections from the

non-QDIO mode SNA Passthru TCP/IP Passthru HPDT MPC

Switch/

Hub/

Router

Server

4/16 Mbps

Token-Ring

Server

DSLw

Router

IP WAN Intranet

10/100 Mbps

Ethernet

4/16 Mbps

Token-Ring

Switch/

Hub/

Router

16/100 Mbps

Token-Ring

Remote Office

4/16 Mbps

Token-Ring

SNA DLSw TCP/IP Native SNA

100 Mbps

Backbone

4/16 Mbps

Token-Ring

Switch/

Hub/

Router

4/16/100 Mbps

Token-Ring

Backbone

IP WAN Intranet

10/100 Mbps

Ethernet

QDIO Mode - TCP/IP

Server

100 Mbps

Token-Ring

4/16 Mbps

Token-Ring

Server

IP Router

Internet or

extranet

Remote Office

4/16 Mbps

Token-Ring

TN3270 browser access to SNA appls. Enterprise Extender for SNA end points TCP/IP applications

remote site to either their TCP/IP or SNA applications on

zSeries and S/390 by conﬁ guring OSA-Express with QDIO

and using either direct TCP/IP access or use appropriate

SNA to IP integration technologies, such as TN3270 Server

and Enterprise Extender for access to SNA applications.

Customers who require the use of native SNA-based con-

nections from the remote site can use a TCP/IP or SNA

transport to the data center and then connect into zSeries

and S/390 using appropriate SNA support on OSA-

Express features conﬁ gured in non-QDIO mode.

LPAR Support of OSA-Express

For z990 customers or customers who use the Processor

Resource/Systems Manager (PR/SM) capabilities IBM

offers the Multiple Image Facility (MIF), allowing the shar-

ing of physical channels by any number of LPARs. Since

a port on an OSA-Express feature operates as a channel,

sharing of an OSA-Express port is done using MIF. The

LPARs are deﬁ ned in the Hardware Conﬁ guration Deﬁ ni-

tion (HCD). Depending upon the feature, and how it is

deﬁ ned, SNA/APPN/HPR and TCP/IP trafﬁ c can ﬂ ow simul-

taneously through any given port.

IPv6 Support

IPv6 requires the use of an OSA-Express adapter running

in QDIO mode and is supported only on OSA-Express

features on zSeries at driver level 3G or above. IPv6 is

supported on OSA-Express for zSeries Fast Ethernet,

1000BASE-T Ethernet and Gigabit Ethernet when running

with Linux on zSeries, z/VM V5.1, and z/OS V1.4 and later.

z/VM V4.4 provided IPv6 support for guest LANs.

more efﬁ cient technique for I/O interruptions designed to reduce path lengths and overhead in both the host operating system and in the adapter. This beneﬁ ts OSAExpress TCP/IP support in both Linux for zSeries and z/VM.

• The z990’s support of virtual machine technology has been enhanced to include a new performance assist for virtualization of adapter interruptions. This new z990 performance assist is available to V=V guests (pageable guests) that support QDIO on z/VM V4.4 and later. The deployment of adapter interruptions improves efﬁ ciency and performance by reducing z/VM Control Program overhead.

Performance enhancements for virtual servers

Two important networking technology advancements are

announced in z/VM V4.4 and Linux on z990:

• The high performance adapter interrupt handling ﬁ rst introduced with HiperSockets is now available for both OSA-Express in QDIO mode (CHPID=OSD) and FICON Express (CHPID=FCP). This advancement provides a

HiperSockets

HiperSockets, a function unique to the zSeries, provides

a “TCP/IP network in the server” that allows high-speed

any-to-any connectivity among virtual servers (TCP/IP

images) and LPARs within a zSeries system without any

physical cabling. HiperSockets decreases network latency

and increases bandwidth between combinations of Linux,

z/OS and z/VM virtual servers. These OS images can be

ﬁ rst level (directly under an LPAR), or second level images

(virtual servers under z/VM).

With new support for up to 16 HiperSockets the z990 pro-

vides four times more HiperSockets, and up to 4,096 TCP/

IP images (stack) connections, which is also four times

more capacity than the z900. The increased HiperSockets

capacity and expanded connectivity provides additional

ﬂ exibility in designing the networking to accommodate con-

solidated and multiple partitioned systems. HiperSockets

can be divided among Logical Channel SubSystems for

separation between various LPARs, while at the same time

a single LPAR could have access to all 16 HiperSockets

the HiperSockets all are assigned to the same LCSS.

A HiperSockets channel also supports spanned chan-

nels in order to communicate between LPARs connected

to different LCSSs. HiperSockets (IQD CHPID type) can

be conﬁ gured to Channel SubSystems and transparently

shared by any or all conﬁ gured LPARs without regard to

the LCSS to

HiperSockets

which the LPAR is conﬁ gured. This means one

could be common to all 30 z990 LPARs. This

support is exclusive to z990. Different HiperSockets can be

used for security (separation of trafﬁ c, no external wire-tap-

ping, monitoring) and performance and management rea-

sons (separate sysplex trafﬁ c, Linux or non-sysplex LPAR

trafﬁ c).

z/VM

LPAR 1

MIF-1 MIF-2 MIF-F MIF-1 MIF-2 MIF-3 MIF-F

CHPID00CHPID

CHPID

PCHID

010B

010D

010C

HiperSockets CHPID 03

z/OS

LPAR

LCSS0 LCSS1

CHPID

PCHID

020A

HiperSockets CHPID 04

CHPID

SPAN

z/VM

LPAR

CHPID

PCHID

0245

0246

HiperSockets CHPID 05

Linux

LPAR

CHPID

PCHID

0248

LinuxLinux

LPAR

CHPID

PCHID

0249

HiperSockets does not use an external network, therefore,

it can free up system and network resources, reducing

attachment cost while improving availability and perfor-

mance. HiperSockets can have signiﬁ cant value in server

consolidation, for example, by connecting multiple Linux

virtual servers under z/VM to z/OS LPARs within the same

z990. Furthermore, HiperSockets can be utilized by TCP/IP

in place of XCF for sysplex connectivity between images

which exist in the same server. Thus z/OS TCP/IP uses

HiperSockets for connectivity between sysplex images in

the same server and uses XCF for connectivity between

images in different servers. Management and administration

cost reductions over existing conﬁ gurations are possible.

HiperSockets acts like any other TCP/IP network interface,

so TCP/IP features like IP Security (IPSec) in Virtual Private

Networks (VPN) and Secure Sockets Layer (SSL) can be

used to provide heightened security for ﬂ ows within the

same CHPID. HiperSockets supports multiple frame sizes,

which is conﬁ gured on a per HiperSockets CHPID basis.

This support gives the user the ﬂ exibility to optimize and

tune each HiperSockets to the predominant trafﬁ c proﬁ le,

for example to distinguish between “high bandwidth”

workloads such as FTP versus lower bandwidth interactive

workloads.

The HiperSockets function provides many possibilities

improved integration between workloads in different

bound only by the combinations of operating systems and

their respective applications. HiperSockets is intended to

provide the fastest zSeries connection between e-business

and Enterprise Resource Planning (ERP) solutions sharing

information while running on the same server. WebSphere

http and Web Application Servers or Apache http servers

can be running in a Linux image (LPAR or z/VM guest) and

will be able to use HiperSockets for very fast TCP/IP trafﬁ c

transfer to a DB2 database server running in a z/OS LPAR.

System performance is optimized because this allows you

to keep your Web and transaction application environ-

ments in close proximity to your data and helps eliminate

any exposure to network related outages, thus improving

availability.

The z/OS HiperSockets Accelerator function can improve

performance and cost efﬁ ciencies when attaching a high

number of TCP/IP images via HiperSockets to a “front end”

z/OS system for shared access to a set of OSA-Express

adapters.

HiperSockets VLAN support in a Linux environment: Virtual

Local Area Networks (VLANs), IEEE standard 802.1q, is

now being offered for HiperSockets in a Linux for zSeries

environment. VLANs can help reduce overhead by allow-

ing networks to be organized for optimum trafﬁ c ﬂ ow; the

for

LPARs,

network is organized by trafﬁ c patterns rather than physi-

cal location. This enhancement permits trafﬁ c to ﬂ ow on a

VLAN connection between applications over HiperSockets

and between applications on HiperSockets connecting to

an OSA-Express Gigabit Ethernet, 1000BASE-T Ethernet,

or Fast Ethernet feature.

HiperSockets broadcast support for IPv4 packets – Linux,

z/OS, z/VM: Internet Protocol Version 4 (IPv4) broadcast

packets are now supported over HiperSockets internal

LANs. TCP/IP applications that support IPv4 broadcast,

such as z/OS OMPROUTE when running Routing Information

Protocol Version 1 (RIPv1), can send and receive broadcast

packets over HiperSockets interfaces. This support is exclu-

sive to z990. Broadcast for IPv4 packets is supported by

Linux for zSeries. Support is planned to be available in z/OS

1.5. Support is also offered in z/VM V4.4 and later.

HiperSockets Network Concentrator

HiperSockets Network Concentrator support, exclusive to

z890 and z990 can simplify network addressing between

HiperSockets and OSA-Express. You can now integrate

HiperSockets-connected operating systems into external

networks, without requiring intervening network routing

overhead, thus helping to increase performance and sim-

plify conﬁ guration. With HiperSockets Network Concentra-

tor support, you can conﬁ gure a special purpose Linux

operating system instance, which can transparently bridge

trafﬁ c between a HiperSockets internal Local Area Network

(LAN) and an external OSA-Express network attachment,

similar to a real Layer 2 switch which bridges between dif-

ferent network segments. This support can

nal HiperSockets network address connection

were directly connected to the external network.

make the inter-

appear as if it

+ 61 hidden pages

IBM z-OS User Manual

Specifications and Main Features

Frequently Asked Questions

User Manual