IBM 325 User Manual

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

Sales Guide

August 2003

IBM ^ 325

Product Overview

CONTENTS

Groundbreaking 32/64-bit performance in a 1U server

[Target Customers: Businesses requiring a High-Performance Computing server with 32-bit
(and, optionally, 64-bit) capabilities in a space-constrained data center, or Scientific/Technical
computing clusters.]

In today’s world, replacing hardware is expensive. The challenge for your customers is to stretch
their IT dollars as far as possible. The 2-way IBM ^ 325 lets them do exactly that by
offering affordable 32-bit computing today with growth to 64-bit computing when needed. It
supports the latest AMD Opteron processors with HyperTransport technology. HyperTransport
technology is a high-speed, point-to-point link between the processor and the memory, graphics
and I/O controllers that provides outstanding performance by bypassing many key bottlenecks
that afflict other processor architectures. Opteron is capable of running both 32-bit and 64-bit
software with blazing performance. This provides customers with the flexibility to start with 32-bit
applications and grow to 64-bit software over time, without having to replace their servers to do

1Product Overview ........

so. In fact, using a 64-bit operating system, both 32-bit and 64-bit applications can be run
concurrently.

2Key Features .............

In addition, the e325 uses leading-edge industry-standard ECC (Error Checking and Correcting)

7Key Options ..............

PC2700 333MHz double data rate (DDR) memory with Chipkill™ technology—for high
performance and reliability. Dual integrated Gigabit Ethernet

8e325 Images ..............

high-speed data transfers and saving valuable adapter slots.

9e325 Specifications .......

All models offer impressive scalability for a 1U server, including dual-processor support, up to

11Selling Features .........

12GB of memory and two high-performance hard disk drives with an internal storage capacity of
up to 293.6GB

12Easy Configurator .......

RAID-1 mirroring. Additional RAID support is available via the optional IBM ServeRAID™ family

14Product Positioning .....

of Ultra160 and Ultra320 SCSI controllers. The ultradense 1U form factor allows businesses to
increase their computing power and spread their workload without outgrowing their current data

14Competitive Analysis ....

center. Up to 42 of these servers can be installed in a single 42U rack, for a total of up to 84
processors, offering your customers tremendous deployment flexibility. Optional Advanced

18For More Information ....

Connectivity Technology (ACT) interconnect cabling technology minimizes cable clutter, reduces

18Legal Information ........

cable ost and minimizes installation time when cabling many e325 servers in a single rack.

Standard in the e325 is an integrated systems management controller, optimized for HPC cluster
manageability, that enables the user to manage and control the server easily—both locally and
remotely. These advanced features help maximize application availability by increasing server
uptime, as do hot-swap redundant HDDs and variable speed fans.

With the inclusion of unique manageability features, such as IBM Director and support for the
optional IBM Remote Supervisor Adapter II (both planned for Q4/03) the e325 is designed for
superior uptime. In addition, IBM Cluster Systems Management can help to reduce the total
cost of ownership for cluster environments by streamlining and simplifying the management of
large numbers of servers from a single point of control.

If your customer needs leading-edge 32-bit performance in a 1U rack-optimized package, with the
added power of 64-bit computing, at a much lower cost than Itanium 2 servers currently
offer—and without replacing hardware, the e325 is the ideal system. To illustrate, in the
SPECfp_rate2000 benchmark

2

(hot-swap SCSI) or 240GB (fixed IDE). The built-in SCSI controller supports

3

, the e325 (in a dual AMD Opteron 246 processor configuration)

1

controllers are standard, providing

1

Actual data transfer speed will vary and is often less than the maximum possible. Gigabit Ethernet transfer speed requires support on
both system and server, and appropriate network infrastructure.

2

GB equals 1,000,000,000 bytes when referring to hard disk drive capacity. Accessible capacity may be less.

3

HP DL360 results: http://www.spec.org/cpu2000/results/res2003q3/cpu2000-20030630-02353.html; HP rx2600 results:

http://www.spec.org/cpu2000/results/res2003q3/cpu2000-20030630-02322.html

; IBM ^ 325 results will be available as of

IBM Confidential 1.

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

posted a floating-point score that was 41% better than that of the Xeon-based HP DL360 G3
server, and at a slightly lower cost. The same e325 score was 63% of that posted by the Itanium
2-based HP rx2600 server, but at one-seventh the cost!

In short: Compared to the x335, the e325 is faster (faster processor for integer operations, faster
DDR memory), has greater memory capacity, and it offers the ability to run both 32-bit and
64-bit applications concurrently. When running a 64-bit operating system, the e325 breaks
through the “4GB barrier” for memory addressing imposed by 32-bit processors.

Key Features

What’s New? (Compared to the x335)

y AMD Opteron processors with HyperTransport technology, integrated memory controller and

1MB of L2 processor cache

y PC2700 333MHz DDR Chipkill ECC memory (vs. PC2100 266MHz DDR Chipkill ECC

memory for the x335)

y 12GB memory capacity (vs. 8MB for the x335)

y Four USB 1.1 ports (2 front, 2 rear)

y Optional USB floppy drive

High-Performance Opteron Processors

The e325 offers a choice of processors:

y AMD Opteron 240, 242 and 246

The integrated memory controller and 1MB of L2 cache run at the full processor clock rate, not the
stepped down rate of the front-side bus (FSB) used in Intel processors. Because of this, the

Opteron 246’s Hyper Transport Tunnel offers peak throughput of 16.0GB per second vs. only

6.4GBps for the (1.5GHz) Itanium’s front-side bus (FSB) — up to 250% faster than Itanium 2, and

a maximum of 4.26GBps for the (3.06GHz) Xeon’s FSB (up to 375% faster than Xeon). Similarly,
the maximum aggregate peak memory throughput of the Opteron in a 2-way configuration is

10.6GBps, or up to 250% higher throughput than the peak rate of 4.26GBps provided by the
533MHz FSB of the Intel Xeon processor used in the x335, and up to 24% more than the 8.5GBps
of the Itanium 2. I/O performance is equally strong, with the Opteron supporting a peak I/O
bandwidth of 12.8GBps, or up to 50% more than the Itanium’s 6.4GBps maximum, and up to 400%
of the 3.2GBps of the Xeon.

The e325 ships with one processor, upgradable to two.

DDR Chipkill ECC Memory

The e325 supports up to 12GB4 of memory in six DIMM sockets. It uses registered PC2700
double data rate (DDR) two-way interleaved memory (operating at 333MHz) for faster access,
and incorporates the same renowned IBM Chipkill ECC technology used in more expensive IBM
servers, for up to 16X better error correction than standard ECC memory.

Memory must be installed in pairs. It is available in 512MB and 1GB (2GB DIMMs are planned).

Large Disk Storage Capacity

The e325 offers a choice of disk storage, supporting up to two internal HDDs:

Hot-swap SCSI

y Ultra320, 10,000 RPMs — 36.4, 73.4 or 146.8GB (293.6GB maximum)

y Ultra320, 15,000 RPMs — 36.4 or 73.4GB (146.8GB)

Fixed IDE

y 7,200 RPM — 40, 60, 80 or 120GB (240GB)

The SCSI drives are hot-swappable, while the IDE drives are not. If the customer needs more
storage space, terabyte capacities are possible with external IBM ServeRAID SCSI or IBM FAStT
Fibre Channel storage solutions. The Ultra320 SCSI hot-swap drives use the Converged Tray for
interchangeability with xSeries systems and FAStT expansion units.

August 11, 2003. Price comparisons using public web prices as of July 28, 2003.

4

In a 2-way configuration, using 2GB DIMMs. See the e325 system specifications table for details.

(ATA-100)

IBM Confidential 2.

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

Ultra320 SCSI Controller

The integrated single-channel Ultra320 SCSI controller (LSI 53C1020) allows speeds of up to
320MB per second
LVD (low-voltage differential) drives with selectable RAID-1 disk mirroring.

Optional IBM ServeRAID controllers increase the RAID support to nine different levels. For
low-cost RAID support of internal IDE drives, use the ATA-133 RAID Controller. Supported RAID
controllers include:

y ServeRAID-6M — Ultra320, 2-channel, 256MB battery-backed cache, 133MHz PCI-X

y ServeRAID-6M — Ultra320, 2-channel, 128MB battery-backed cache, 133MHz PCI-X

y ServeRAID-4Lx — Ultra160, 1-channel, 32MB cache, 66MHz PCI

y ATA-133 RAID Controller (IDE) — ATA-133, ATA-100, ATA-66, 2-channel, half-length

The ServeRAID-4 and ServeRAID-6M controllers support external SCSI expansion via the IBM
EXP300 Storage Expansion Unit. Each EXP300 adds up to 2.055TB of external Ultra160 SCSI

storage capacity.

The RAID subsystem is closely linked with IBM Director to allow proactive monitoring of the disk
subsystem. This feature is used to generate alerts in the event of an impending disk failure or an
actual disk failure via Predictive Failure Analysis

5

across the SCSI bus and supports two internal Ultra320 or Ultra160 SCSI

®

(PFA).

Dual-Channel Gigabit Ethernet Controller

The e325 includes a two-channel integrated Broadcom BCM5704 Gigabit Ethernet controller for
up to 10X higher maximum throughput than the dual 10/100 Ethernet controllers used in some
systems. Having integrated controllers saves PCI slots. Optional PCI adapters offering failover
and load balancing between them are available for better throughput and system availability.

What’s Hot?

64-bit High-Speed PCI-X Adapter Slots

The e325 has one full-length and one half-length adapter slot, capable of supporting 100MHz
64-bit and 32-bit PCI-X adapters and/or older 64-bit and 32-bit 66MHz and 33MHz PCI adapters.
Because the Ultra320 SCSI, dual Gigabit Ethernet and systems management controllers are
integrated onto the system board, the two available high-speed adapter slots offer customers a
wide degree of latitude in expansion options.

Other New Features

y Four USB 1.1 ports — The flexibility to add more external devices, with two ports in the front

and two in the rear.

y High-resolution video — Up to 1600x1200 resolution, with a color depth of 24 bits at 85Hz

refresh rate, using the ATI Rage XL SVGA video controller.

y USB-attached floppy drive support — Optional drive can be relocated to share among

multiple systems.

y Toolless features — No tools are required for rack installation, cover removal, adapter

retention, power supply replacement or HDD installation/removal.

y Dual processor capability providing symmetrical multiprocessing (SMP) with HyperTransport

technology and integrated memory controller

y Systems management capabilities appropriate for the HPC environment

y Extensive systems management and support features

Dual SMP Processors with HyperTransport Technology

The e325 supports up to two high-performance AMD Opteron processors, allowing customers to
upgrade to a second processor as business needs require.

HyperTransport technology is a high-speed, low-latency, point-to-point link, designed to increase
the communication speed between circuits. It eliminates a key bottleneck currently caused by the

5

Data transfer rates depend on many factors and are often less than the maximum possible.

IBM Confidential 3.

f

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

stepped-down front side bus (FSB) used in other system architectures, by accessing memory at
the same speed as the processor. This permits the fastest possible access to CPU and memory.

Opteron offers outstanding 32-bit performance using today’s operating systems, and an upgrade
path to low-cost 64-bit computing with tomorrow’s 64-bit OSes. With a 64-bit OS, Opteron is even
capable of running both 64-bit and 32-bit applications concurrently.

Go to http://amd.com

for more information on HyperTransport technology.

Rack Cable Management and KVM Console Switching

IBM Advanced Connectivity Technology (ACT) is an optional feature that offers many of the same
advantages as IBM C2T™ Interconnect cabling (used with x330/x335 servers), plus a few others,
across the entire xSeries product line. So now customers can interconnect all of their xSeries and
Netfinity servers with one smart cabling architecture. Like C2T Interconnect technology, ACT
cabling eliminates the need for one-to-one direct connections between each server and a KVM
switch.

Here’s how it works: A USB Conversion Option (UCO) plugs into a USB port of an e325 server
and converts the KVM signal to standard CAT5 network cabling signals. An inexpensive CAT5
cable (with RJ-45 connectors on each end) then connects the first server to a UCO or a KCO
(KVM Conversion Option, used by xSeries servers) in the next server in the rack. This allows

Short KCO

customers to daisy-chain servers much like the C2T Interconnect feature. Because CAT5 cabling
is industry-standard, customers can assemble cables to fit
16 servers to be chained together for common management. If customers have x330/x335
servers in the rack, there is a special 110mm (4”) C2T Conversion Option (CCO) that takes the
C2T output and converts it to CAT5 signals for input to one of the CAT5-based switches.

There are two CAT5-based console switches available to manage the servers via the UCO, KCO
and CCO kits: Local Console Manager (LCM), a 4-port CAT5 console switch and Remote
Console Manager (RCM), a 16-port switch. Each converts the CAT5 signals back to KVM signals
for input to a management station.

Using the LCM, up to 16 servers can easily be daisy-chained together (using 16 KCOs) to each o
its four CAT5 inputs, enabling the management of up to 64 servers. It provides a single user with
local access over all connected systems. The RCM can handle up to 256 servers using any
combination of UCO, KCO and CCO kits, and supports a single local user and up to two remote
users simultaneously.

Both the LCM and the RCM can accept legacy KVM switch feeds by equipping the switch with
one of the KCO options. This makes it easy to incorporate ACT into a traditional KVM
environment. The illustration below shows a sample ACT configuration:

6

on-site. The UCO/KCO allows up to

Remote Console
Manager

Legacy NetBAY KVM switch

Up to 42 chained
x330/x335 servers per
port via CCO

Note: ARI (Analog Rack
Interface) = CAT5 port

6

To achieve the highest video quality (1600x1280 resolution) for local management, the distance from the LCM or RCM to the
farthest server on a chain should not exceed 15m (50 ft). For remote access at 1280x1024 resolution, total CAT5 cabling in any
chain should be limited to about 10m (32 ft). Longer cables can be used, however video degradation is possible beyond these
lengths. For more on ACT cabling, see the ACT Decision Makers’ Guide at

ibm.com/support/docview.wss?uid=psg1MIGR-46120.

IBM Confidential 4.

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

Like C2T Interconnect cabling, ACT reduces clutter, simplifies cable management, and reduces
cabling costs. Also, having fewer cables improves rack airflow and reduces heat buildup. Plus,
while typical KVM cabling limits the management to perhaps 16 servers per switch, ACT can
support up to 256 servers using only one RCM. The reduced switch count means that more
space is freed up in the side-pocket space of customers’ racks; switch and PDU costs should be
greatly reduced; server administration is simplified and there are fewer points of potential failure.

Variable Speed Fans

Eight strategically located fans provide effective system cooling for the e325, from front to back,

throughout the system. This cooling scheme is important because newer, more powerful
processors generate a significant amount of heat, and heat needs to be controlled for the system
to function properly. The e325 has four system fans, one I/O fan and three fans integrated into
the power supply.

The fans automatically adjust speeds in response to changing thermal requirements. When the
temperature inside the server increases, the fans speed up to maintain the proper ambient
temperature. When the temperature returns to a normal operating level, the fans return to their
default speed. Why not simply run the fans at 100% capacity all the time? For several good
reasons: to reduce the ambient noise, reduce the wear-and-tear on the fans and reduce the
server power draw. The reduction in ambient noise and power draw may be relatively minor for a
single server, but put dozens or hundreds in a data center and it makes a big difference!

Drive Bays

The e325 contains three bays. Two bays are hot-swap-capable (when used with SCSI hard disk
drives). This enables up to two slim-line (1") 3.5” drives to be installed for a total hot-swap
Ultra320 SCSI disk capacity of 293.6GB. (Two fixed IDE drives can be used instead, totaling up
to 240GB.) The bays accept drives that use the Converged Tray, for compatibility with other
xSeries servers. A 24X
bay. (DVD, CD-RW and DVD/CD-RW combo drives are available as options to replace the
CD-ROM drive.) If a 1.44MB diskette drive is required, one is available as an option to attach
externally to a USB port.

If the customer requires more than two 3.5" HDDs, an optional ServeRAID SCSI controller, or an
IBM FAStT Fibre Channel adapter can be installed, along with external storage units.

7

speed IDE CD-ROM drive is standard in the other (ultraslim half-inch)

Extensive Systems Management Capabilities

The e325 has a high level of systems management capability, including an integrated systems
management controller, Alert Standard Format, Automatic Server Restart, Wake-on-LAN support,
PXE support, Predictive Failure Analysis, IBM Director and planned support for the optional
Remote Supervisor Adapter II. This allows high levels of manageability for customers, without
their having to purchase additional adapters or options.

The integrated systems management controller provides local and remote monitoring of the
server, including the monitoring of system voltages, temperatures, fans, etc. It also controls the
fan speed and the diagnostic LEDs. It features text console redirection over shared NIC or serial
port, out-of-band LAN-based management, remote out-of-band alerts, support for command-line
interface over shared NIC or serial port, remote access security, out-of-band environmental
monitoring and alerting, secure remote power control and system reset over shared NIC or serial
port, and others.

Alert Standard Format (ASF) firmware alerts IBM Director to anomalous environmental factors,
such as voltage and thermal conditions—even if the server has failed.

Automatic Server Restart (ASR) helps reduce downtime by restarting the server automatically in
the event of a system lockup. ASR technology is a combination of hardware circuitry tied into the
server’s system reset function and a device driver. As long as the server continues running, the
ASR watchdog timer will keep being reset, but if the operating system crashes or the hardware
freezes somehow the ASR software will be unable to reset the hardware timer. If the timer is not
reset within five minutes, it automatically triggers the ASR hardware, which immediately restarts
the server (and logs an ASR event with IBM Director).
more than five minutes passing before the server is restarted.

Wake on LAN permits the e325 to be remotely powered on if it has been shut off. Once powered
up the server can be controlled across the network, using the Preboot Execution Environment
(PXE 2.0).

7

Variable read rate. Actual playback speed varies and is often less than the maximum possible.

These features are designed to prevent

IBM Confidential 5.

Outstanding 32/64-bit price/performance in a rack-dense 1U, 2-way server for HPC
applications

Like Wake on LAN, PXE is system firmware. It allows systems management software to take
control of a system before the BIOS, operating system or applications are loaded (using Wake on
LAN/PXE) and lets an administrator perform many low-level tasks remotely that would otherwise
require a visit to each system. These tasks may include such things as formatting a hard disk
drive, updating system firmware (BIOS or ISMP, for example), or deploying a Windows or Linux
operating system.

Predictive Failure Analysis (PFA)

hard disk drives as much as 48 hours before actual failure, and alert the administrator through

IBM Director. This helps give customers the ability to replace the failing component before it fails,
resulting in increased uptime.

IBM Director 4.1x (planned for Q4/03), software for advanced workgroup management, is
offered at no additional cost for the e325. IBM Director comes with a portfolio of tools for server
management and increased availability. IBM Director provides a single uniform graphical
interface for all systems management functions.

The combination of IBM Director and the integrated systems management controller enables the
customer to customize thresholds and monitor system components (for things like temperature,
voltage regulation, etc.) to help maximize uptime.

Plans call for the e325 to also support (effective Q4/03) an optional Remote Supervisor
Adapter II for additional systems management capabilities, including:

y Web-based out-of-band control

y Windows “blue screen” capture

y Remote redirection of graphics, keyboard and mouse

y The ability to manage a rack of up to 24 servers from a single Ethernet or serial connection

using one Remote Supervisor Adapter.

Note: This information represents IBM's current intentions, goals and objectives, and is subject to
change or withdrawal without additional or prior notice.

is designed to allow the e325 to detect impending failure of

Extensive System Support Features

The IBM services and technical support portfolio provides world-class, consistent, high-quality
service and support. The e325 server offers a number of tools and services designed to make
ownership a positive experience. From the start, IBM programs make it easier for customers to
plan for, configure and purchase IBM ^ systems, get them running and keep them
running long-term. These programs and features include IBM ServerProven
Systems Management, the IBM ^ xSeries and Netfinity Rack Configuration Tool, IBM
ServerGuide™, Product Customization Services and extensive technical support offerings.

The IBM ServerProven program provides the confidence that specific options and operating
systems have been tested on the e325 and are officially supported to work together. It is updated
frequently with that the latest compatibility information.

IBM Cluster Systems Management (CSM) can help to reduce the total cost of ownership for
cluster environments by streamlining and simplifying the management of large numbers of
servers from a single point of control. This is of particular value in large data centers or for
customers who are running complex high-performance workloads, where managers are
constantly challenged to meet ever-changing business demands and improve service levels at
the same time that IT budgets are shrinking. As customers evolve from using clusters for
single-purpose workloads to employing cluster technology in grids or utility models, CSM can
provide a path for growth and enhanced value.

Customers who have existing AIX-based cluster systems can leverage those skills to manage
their Linux clusters. System administrators can automate problem determination and recovery,
automate repetitive installation and configuration tasks, and monitor and report health information
and resource utilization.

CSM is intended primarily for managing clusters of Linux-based xSeries servers or AIX-based

™

pSeries
with nonclustered Linux servers. Typically, customers would use IBM Director to manage
all-Windows or a mix of Windows and Linux-based xSeries systems in a nonclustered
environment. CSM would be the preferred choice for Linux-only nonclustered xSeries servers or
clustered Linux/AIX systems (both xSeries and pSeries)—especially in a high-performance
computing (HPC) environment.

IBM ^ xSeries and Netfinity Rack Configurator is a downloadable tool that simplifies
the often complex chore of configuring a full rack of e325, BladeCenter and xSeries servers and

servers, or for heterogeneous clusters combining the two, although it also works well

®

, IBM Cluster

IBM Confidential 6.