IBM eServer zSeries 900 Technical Guide

2 IBM eServer zSeries 900 Technical Guide

1.1 Introduction

The IBM zSeries 900 (z900) family of servers offers 42 air-cooled models, from one-way to

16-way, utilizing up to 20 processing units. These servers can be configured in numerous

ways to offer outstanding flexibility and speed deployment of e-business solutions. Each z900

server can operate independently or as part of a Parallel Sysplex® cluster of servers. In

addition to supporting z/OS™, the z900 can host from tens to hundreds of Linux® images

running identical or different applications in parallel, based on z/VM™ virtualization

technology.

The z900 family is based on the IBM z/Architecture™, which supports a new standard of

performance and integration by expanding on the balanced system approach of the IBM

S/390® architecture.

The z900 is designed to eliminate bottlenecks associated with the lack of addressable

memory through its virtually unlimited 64-bit addressing capability, providing plenty of

“headroom” for unpredictable workloads and growing enterprise applications.

Robust network and I/O subsystem connectivity provides a balanced system design. High

speed interconnects for TCP/IP communication, known as HiperSockets™, let TCP/IP traffic

travel between partitions at memory speed, rather than network speed. A high performance

Gigabit Ethernet feature is one of the first in the industry capable of achieving line speed: one

Gigabit per second. Furthermore, the availability of native FIber CONnectivity (FICON™)

devices, Fibre Channel Protocol (FCP) channels, 2 Gigabit/sec links, and FICON Support of

Cascaded Switches can increase I/O performance, consolidate channel configuration, and

help reduce total cost of ownership. The total result is ultra high speed communications within

the server, between servers, to devices, and out to users, allowing greater integration

between traditional and Web applications to maximize e-business effectiveness.

zSeries 900 has an enhanced I/O subsystem. The I/O subsystem includes Dynamic Channel

Path Id (CHPID) Management (DCM) and channel CHPID assignment. These two functions

increase the number of CHPIDs that can be used for I/O connectivity. In the servers prior to

the z900 it was not always possible to use the full range (256) of CHPIDs; for example, the

installation of an OSA-2 required the allocation of 4 CHPIDs, one of which was usable and the

remaining 3 “blocked” and not available for use. This is no longer the case with z900. The

exploitation of these functions allows the full use of the bandwidth available for 256 channels

in the z900. The subchannel addresses have been increased to 512 K for the system and

63 K for an LPAR.

Within the z900 the number of FICON channels, operating either in FICON or FCP modes,

has been increased to 96, giving the z900 three times the concurrent I/O capability of a fully

configured IBM 9672 G6 Server. Fewer FICON channels are required to provide the same

bandwidth as ESCON, reducing channel connections and thus reducing I/O management

complexity. FICON also addresses the architectural implementation constraints of ESCON.

For example, the number of devices per channel increases from 1 K to 16 K.

The z900 has improved the coupling efficiency when configured in a Parallel Sysplex by

increasing both long distance InterSystem Coupling (ISC) channels and short distance

Integrated Cluster Bus (ICB) bandwidth, as well as improving the message passing protocols.

The z900 family of servers also automatically directs resources to priority work through

Intelligent Resource Director (IRD). The z900 IRD combines the strengths of three key

technologies: z/OS Workload Manager (WLM), Logical Partitioning, and Parallel Sysplex

clustering.