2 system z ibm system z hardware

1

© 2006 IBM Corporation IBM Systems

Linux on IBM System z9 and zSeriesSystem Hardware Overview

Linux on System z WorkshopNew Technology CenterIBM France – Montpellier PSSC

2 08/11/2007IBM System z Hardware IBM Systems

Press Conference

7th April 1964, Poughkeepsie NY

• A new generation of electronic computing equipment was introduced today by International Business Machines Corporation. IBM Board Chairman Thomas J. Watson Jr. called the event the most important product announcement in the company's history.

• The new equipment is known as the IBM System/360.

• "System/360 represents a sharp departure from concepts of the past in designing and building computers. It is the product of an international effort in IBM's laboratories and plants and is the first time IBM has redesigned the basic internal architecture of its computers in a decade. The result will be more computer productivity at lower cost than ever before. This is the beginning of a new generation - - not only of computers - - but of their application in business, science and government."

2


�S/390 is dead or terminally ill

ƒthe aging user base is deserting the MVS ship and moving to UNIX or NT

�S/390 is expensive to buy and expensive to run

ƒUNIX or Wintel is many times cheaper

�S/390 requires a huge Datacenter/Glasshouse

ƒbig machines, water cooling and huge electricity bills

�S/390 cannot run new eBusiness or ERP workloads

ƒit only runs Batch or Green-screen type work

9021-9X2 - 470 MIPS in 1994

Water Cooling equipmentas used 1970s to 1990s

.... but that was before IBM "Downsized the Mainfra me"

1990 – The Mainframe is dead – or so they said


Bipolar, Water Cooled Mainframes transition to smal ler CMOS technology

�CMOS significantly reduces cost, size & running costs

ƒSMP goes to 10-way, 12-way and 16-way systems

ƒParallel Sysplex clustering introduced in 1994 to enable shared data

�1996 - fully integrated UNIX environment on MVS/ESA

ƒXPG4 and UNIX/95 branding - enables new workloads

�Over time support is added for eBusiness workloads

ƒIEEE Floating point and Integrated Crypto engines - good performance

ƒOSA and Gb Ethernet support - full conectivity to Internet

�1999 adds FICON channels - Fibre Channel stage-II

ƒhigher speed, longer distance - more D/R options, connection to 'Open Systems I/O'

�2000 - z900 and z/OS - new platform support for eBusiness

ƒSharing systems resources in line with business goals

�2000 introduces Linux on zSeries

ƒNew options for fast start eBusiness applications

ƒIntegration with z/OS via HiperSockets

The technology change …

3


IBM System z9 and eServer zSeriesThe incorporation of classical mainframe strengths

�Reliability, Availability, Serviceability

�Security (cryptographic HW and more ...)

�Scalability:�Scale up on demand�Scale out on demand

Based on:

ƒ Architecture� 64bit, QDIO, ...

ƒ Technology, Packaging� MultiChipModule, ...

ƒ Implementation� Internal and external bandwidth� 'Storage Hierarchy': L1, L2,...

ƒ Virtualization: 'sharing physical resources'� CPU, memory, ...� I/O, network, (LAN)

► Most advanced On Demand Technologies� Redundant resources, hot-pluggable IO, …, OOCoD


Short history

� 1998-06-23 9672-nn6 G5

� 1999-05-03 9672-nn7 G6

� 2000-10-03 2064-1nn z900

� 2002-02-19 2066 z800

(Linux only model 2002-01-29)

� 2002-04-30 2064-2Cn z900 Turbo

� 2003-05-13 2084 z990

� 2004-04-07 2086 z890

� 2005-07-26 2094 System z9 EC

� 2006-04-27 2096 System z9 BC

4


IBM eServer zSeries 890 z890 (2086)IBM eServer zSeries 990

z990 (2084)

� Announced 5/03 – first zSeries Superscalar Server with up to 48 PUs

� 4 models – Up to 32-way

� Specialty Engines► CP, IFL, ICF, zAAP

�On Demand Capabilities ► CUoD, CIU, CBU, On/Off CoD

�Memory – up to 256 GB

�Channels► Four LCSSs ► Up to 1024 ESCON ® channels► Up to 240 FICON Express2 channels► Token-Ring, GbE, 1000BASE-T Ethernet► Coupling Links

�Crypto Express2

� Parallel Sysplex clustering

�HiperSockets ™ – up to 16

�Up to 30 logical partitions

�Operating Systems► z/OS, z/VM®, VSE/ESA™, z/VSE™, TPF,

z/TPF, Linux ® on zSeries

� Announced 4/04 – zSeries Superscalar Server with 5 PUs

� 1 model – Up to 4-way► 28 capacity settings

� Specialty Engines► CP, IFL, ICF, zAAP



�Channel► Two LCSSs► Up to 420 ESCON channels► Up to 80 FICON Express2 channels► Networking Adapters (OSA)► Coupling Links

�Crypto Express2


�HiperSockets – up to 16


�Operating Systems► z/OS, z/OS.e, z/VM, VSE/ESA, z/VSE,

TPF, z/TPF, Linux on zSeries

IBM System z9 (z9 EC) (2094)

� Announced 7/05 - Superscalar Server with up to 64 PUs


�Granular Offerings for up to 8 CPs

� Specialty Engines► CP, IFL, ICF, zAAP, zIIP



�Channels► Four LCSSs► Multiple Subchannel Sets► MIDAW facility► 63.75 subchannels► Up to 1024 ESCON channels► Up to 336 FICON channels► Enhanced FICON Express2 and 4► 10 GbE, GbE, 1000BASE-T► Coupling Links

�Configurable Crypto Express2




� Enhanced Availability

�Operating Systems► z/OS, z/VM, VSE/ESA, z/VSE, TPF,

z/TPF, Linux on System z9

IBM System z9 (z9 BC) (2096)

� Announced 4/06 - Superscalar Server with 8 PUs


�High levels of Granularity available► 73 Capacity Indicators

� Specialty Engines► CP, IFL, ICF, zAAP, zIIP



�Channels► Two LCSSs► Multiple Subchannel Sets► MIDAW facility► 63.75 subchannels► Up to 420 ESCON channels► Up to 112 FICON channels► Enhanced FICON Express2 4 Gbps► 10 GbE, GbE, 1000BASE-T► Coupling Links

�Configurable Crypto Express2




� Enhanced Availability

�Operating Systems► z/OS, z/OS.ez/VM, VSE/ESA, z/VSE, TPF,

z/TPF, Linux on System z9

IBM System z family


z900z/OS 1.4

z990z/OS 1.4

z9 ECz/OS 1.6

20 engines16-way

48 engines32-way

64 engines54-way

Maximum ITRsMinimum ITRs

Mainframes Continues to Scale with System z9

� Each new range is designed to deliver:

► New function

► Improved performance

► Improved availability and serviceability

► Field upgrades to help protect customer investment

► Better price/performance

5


z9 BC – Delivering increased capacity and performance

z9 BC

Capacity

z890 z800

Pro

cess

or U

nits

34% improvement

75%

mor

e

� Improved I/O Performance► 40% more FICON channels – up to 112►Up to 170% more bandwidth than z890►Can improve FICON performance with

Modified Indirect Data Address Word (MIDAW) facility

►Double the FICON concurrent I/O operations from 32 to 64 on FICON channel

►Multiple Subchannel Sets (MSS) for an increased number of logical volumes

�Greater granularity and scalability► Two models with one machine type (2096)

●1 to 4-way high performance server standard engines●Entry model with 1 to 3-way standard engines●Up to a 7-way with specialty engines

► 73 capacity settings for a 260% increase in flexibility over z890

► Delivers up to 34% more capacity with the same low entry point as the z890

► Up to 34% hardware performance improvement for Linux (IFLs), Java (zAAPs) and Internal Coupling Facilities (ICFs)

► New zIIP for data serving workloads ► Double the memory – up to 64 GB per server


� z9 EC can provide increased capacity in a single foo tprint►Faster uniprocessor performance than the z990

►Additional engines ► Increased capacity on specialty engines for consolidation

� On demand advancements for unplanned and planned ou tage reduction

� May boost performance of new workloads and security

z9 EC Product Positioning

z9 EC

MIPS capacity in a single footprint

z990

z900

6


�Machine Type►2096

�Model►R07

�Processor Units (PUs)►8 PUs per System►1 SAP per book, standard►No dedicated spares ►7 PUs available for characterization

● 1 to 3 Central Processors (CPs)● Integrated Facility for Linux (IFLs), Internal Coupling

Facility (ICFs), System z Application Assist Processors (zAAPs), System z9 Integrated Information Processors (zIIPs), optional System Assist Processors (SAPs)

►Up to 15 LPARs�Memory

►Minimum of 8 GB►Up to 64 GB per System

● 8 GB increments� I/O

►Up to 16 STIs per System @ 2.7 GB/s each►Total system I/O bandwidth capability of 43.2 GB►Up to 2 Logical Channel Subsystem (LCSS)►Up to a maximum of 4 I/O Domains

● Up to 240 channels – dependent on Channel types

z9 BC Model R07



�Model►S07

�Processor Units (PUs)►8 PUs per System►1 SAP per book, standard►No dedicated spares ►7 PUs available for characterization

● 0 to 4 Central Processors (CPs)● Integrated Facility for Linux (IFLs), Internal Coupling Facility

(ICFs), System z Application Assist Processors (zAAPs), System z9 Integrated Information Processor (zIIP), optional System Assist Processors (SAPs)

● Can have an IFL-only System►Up to 30 LPARs

�Memory►Minimum of 8 GB►Up to 64 GB per System


►Up to 16 STIs per System @ 2.7 GB/s each►Total system I/O bandwidth capability of 43.2 GB►Up to 2 Logical Channel Subsystems (LCSSs)►Up to a maximum of 7 I/O Domains

● Up to 420 channels – dependent on Channel types

z9 BC Model S07

7



�5 Models►S08, S18, S28, S38 and S54

�Processor Units (PUs)►12 (16 for Model S54) PUs per book►2 SAPs per book, standard►2 spares per server►8, 18, 28, 38 or 54 PUs available for characterization

● Central Processors (CPs), Integrated Facility for Linux (IFLs), Internal Coupling Facility (ICFs), System z Application Assist Processors (zAAPs), System z9 Integrated Information Processor (zIIP), optional System Assist Processors (SAPs)

�Memory►Minimum of 16 GB►Up to 128 GB per book►Up to 512 GB for System


►Up to 16 STIs per book @ 2.7 GB/s each►Total system I/O bandwidth capability of 172.8 GB*►Up to 4 Logical Channel Subsystems (LCSSs)

* z9 EC exploits a subset of its designed I/O capability

z9 EC Overview


I/O Cage

PowerSupplies

InternalBatteries(optional)

Single Processor Book and Memory

CEC Cage

Support Elements(gate with Laptops swung open)

STI Connectors

Front View

A Frame

Fiber Quick Connect Feature

(optional)

z9 BC – Under the covers

8


HybridCooling

Processor Books and Memory

CEC Cage

STI cables

SupportElements

3x I/Ocages

PowerSupplies

InternalBatteries(optional)

Front View

Fiber Quick Connect Feature

(optional)

z9 EC – Under the covers (Model S38 or S54)


MSC MSCPU PU

PU

PU

PU PU

PU

PUSD SD

SD SD

SCCLK

z9 BC 8-way MCM

� Advanced 95mm x 95mm MCM► 102 Glass Ceramic layers► 16 chip sites, 217 capacitors► 0.545 km of internal wire

� CMOS 10Ks0 chip Technology►PU, SC, SD and MSC chips►Copper interconnections, 10 copper layers►8 PU chips/MCM

●15.78 mm x 11.84 mm●L1 cache/PU

– 256 KB I-cache●121 million transistors/chip

– 256 KB D-cache●0.7 ns Cycle Time

►4 System Data (SD) cache chips/MCM●15.66 mm x 15.40mm●L2 cache per Book

– 660 million transistors/chip– 40 MB

►One Storage Control (SC) chip●16.41mm x 16.41mm●162 million transistors●L2 cache crosspoint switch●L2 access rings to/from other MCMs

►Two Memory Storage Control (MSC) chips●14.31 mm x 14.31 mm●24 million transistors/chip●Memory cards (L3) interface to L2●L2 access to/from MBAs (off MCM)

►One Clock (CLK) chip - CMOS 8S●Clock and ETR Receiver

9


MSC MSCPU PU

PU

PU

PU PU

PU

PUSD SD

SD SD

SCCLK

z9 EC 12-way MCM

� Advanced 95mm x 95mm MCM► 102 Glass Ceramic layers► 16 chip sites, 217 capacitors► 0.545 km of internal wire

� CMOS 10Ks0 chip Technology►PU, SC, SD and MSC chips►Copper interconnections, 10 copper layers►8 PU chips/MCM

●121 million transistors/chip●L1 cache/PU●15.78 mm x 11.84 mm

– 256 KB I-cache– 256 KB D-cache

●0.58 ns Cycle Time►4 System Data (SD) cache chips/MCM

●15.66 mm x 15.40mm●L2 cache per Book

– 660 million transistors/chip– 40 MB

►One Storage Control (SC) chip●16.41mm x 16.41mm●162 million transistors●L2 cache crosspoint switch●L2 access rings to/from other MCMs

►Two Memory Storage Control (MSC) chips●14.31 mm x 14.31 mm●24 million transistors/chip●Memory cards (L3) interface to L2●L2 access to/from MBAs (off MCM)

►One Clock (CLK) chip - CMOS 8S●Clock and ETR Receiver


Note : 1. Concept Illustration only - not to scale 2. 4 or 8 pluggable Memory Cards 3. Each MBA fanout card is hot-pluggable and has 2 STIs

Front View Side View

"D6"

"D4" "D5"

"F1" "D1" "D2"

"D3"

"D7" "D8"

MSC

MSC

CP 2

CP CP

HitachiCP HitachiCP

SD

SD SD

SD

SC

Memory Cards up to 64 GB (z9BC)128GB (z9EC)

MCMUp to 8Hot pluggable MBA/STI fanout cards

z9 BC Processor Book Layout

10


�PU 4 always is SAP, each PU is single core and Crypto Assist

�L2 contains Storage Controller and Storage Data Chips

�Book Package contains PU, Storage Controllers, L2 and L3 Caches, 2.7 GB/sec Self-Timed Interfaces

�Seven of 16 STIs can attach to the I/O cage. Rest can be used for ICB-4s

I/O Cage with 7 I/O domains

ESCON 16 PortFICON Express2 4 portFICON Express4 2 or 4 portOSA-Express2 1 or 2 portCrypto Express2 2 CoprocessorISC-3 1-4 Port

Up to 28 I/O Adapters:

ICB-4 Links2 Gbyte/s

ETR-Links

STI- Links2.7 Gbyte/s

MSC 0 MSC 1

PU0

L1

PU2

L1

PU4

L1

PU5

L1

PU7

L1

MBA MBA

Up to 64 GB Memory(L3 Cache) CLK

PU6

L1

PU3

L1

PU1

L1

1 Cache Control Chip + 2 Cache Data Chips(40 MB Shared L2 Cache)

MBA MBAMBA MBAMBA MBA

Note: STI connections to the I/O cage are NOT representative of an actual configuration.

z9 BC 8 PU Logical Structure


Processor

Card

Slot 1Book 3

Slot 4Book 2

Slot 2Book 0

Slot 3Book 1

Book 0

Book 1

Book 2

Book 3

CEC Board backplane

MCMProcessor

Card

MCMProcessor

Card

MCMProcessor

Card

MCM

� The ring structure consists of two rings (one running clockwise, the other counterclockwise)

� In a two or three Book configuration, Jumper Book(s) (installed in the CEC cage) complete the ring► Jumper Books are not needed for a single-Book

configuration � Books may be able to be inserted into or removed

from the ring nondisruptively *►May allow Concurrent book add for model

upgrade►Enhanced book availability to return a book after

removal for upgrade or repair

MB

AF

anoutM

emory

MB

AF

anoutM

emory

MB

AF

anoutM

emory

MB

AF

anoutM

emory

* Customer pre-planning required, may require acquisition of additional hardware resources

z9 EC Model S38 – Communication Ring Structure

11


Internal Coupling Facility (ICF) 1997

Integrated Facility for Linux

(IFL) 2001

IBM System z9 Integrated Information Processor (zIIP) 2006

IBM System z Application Assist Processor (zAAP)

2004

Building on a strong track record of

technology innovation with specialty

engines, IBM introduces the System

z9 Integrated Information Processor

�Support for new workloads and open standards

�Designed to help improve resource optimization for eligible data workloads within the enterprise

�Centralized data sharing across mainframes

�Designed to help improve resource optimization for z/OS Java™

technology-based workloads

Technology evolution with specialty engines


� Processor Units (PUs) that can be ordered on System z9

►Central Processor (CP)●Provides processing capacity for z/Architecture™ and ESA/390 instruction sets●Runs z/OS, z/VM, VSE/ESA, z/VSE, TPF4, z/TPF, Linux for System z and Linux under

z/VM or Coupling Facility● z9 BC has Capacity Marker features NOT Unassigned CP features

►IBM System z Application Assist Processor (zAAP)●Under z/OS, the Java Virtual Machine (JVM) assists with Java processing to a zAAP

►IBM System z9 Integrated Information Processor (zIIP) – when available●Provides processing capacity for selected workloads e.g., DB2 for z/OS V8 workloads

executing in SRB mode►Integrated Facility for Linux (IFL)●Provides additional processing capacity for Linux workloads

►Internal Coupling Facility (ICF)●Provides additional processing capacity for the execution of the Coupling Facility Control

Code (CFCC) in a CF LPAR►Optional System Assist Processors (SAP)●SAP manages the start and ending of I/O operations for all Logical Partitions and all

attached I/O

System z9 PU Characterization

12


Input/Output hardware

zSer

ies/

z9 p

roce

ssor

ESCONChannel

I/OControl

Unit

DeviceDevice

ESCONChannel

ESCONChannel ESCON

Director

I/OControl

UnitI/O

Control Unit

I/OControl

UnitI/OControl

Unit

Device

Device

Device

Device

Device

DeviceDevice

FICON Express2Channel

I/OControl

UnitFICONswitch

I/OControl

UnitI/O

Control Unit

DeviceDevice

DeviceDevice

FICON Express2Channel

OSA Express2Adapter LAN

� Terms: Channel, ESCON, FICON, OSA, Hipersockets

� Up to 1024 channels, total bandwidth > 170GB/secHipersocket

FCP (FICON channel)

FCPswitch SCSI Device

SCSI Device

Storage Ctrl UnitSAN Fabric


Horizontal scaling & high availability within the data center

� Parallel Sysplex

► Loosely coupled multiprocessing

► Requires:

► Hardware/software combination● Data sharing● Locking● Cross-system workload dispatching● Synchronization of time for logging, etc.● High-speed system coupling

► Hardware:● Coupling Facility

– Integrated Cluster Bus and ISC to provide high-spee d links to CF

● Sysplex Timer – Time Of Day clock synchronization

► Implemented in z/OS* and subsystems● Workload Manager in z/OS● Compatibility and exploitation in software subsyste ms,

including IMS*, VSAM*, RACF*, VTAM*, JES2* , etc.

121

2

3

45

67

8

9

10

11

CouplingFacility

Shared data

Sysplex Timers

ESCON/FICON*zSeries

IBM System z

1 21

2

3

4

5

6

7

8

9

1 0

1 1

13


Hardware management/maintenance - Support Element and HMC

� Two Support Elements (SE) are connected to two internal Ethernet LANs

� There is a crossover capability between the LANs, so that both SEs have access to both internal LANs

� With a second Ethernet adapter, the SEs are also connected to an external Ethernet LAN

� The Hardware Management Console (HMC) is connected to the external LAN

DCADCA DCADCADCADCADCADCA

Bulk power Bulk powerFSP = Flexible Support Processor

Ethernet “hub” 1

Ethernet “hub” 2

book 0 book 1 I/O cage 3I/O cage 2

PrimarySupportElement

AlternateSupport Element

HMC

externalLAN

FSP FSP FSP FSPFSP FSP

FSPFSP

FSP FSP


� z/OS► The flagship zSeries/z9 operating system

► Ultra-high reliability – over 50% of the OS is dedicated to error recovery!

� z/VM► Run any zSeries/z9 operating system as a “guest” – Large scale virtualization

► Sometimes used for application development

� Linux for IBM System z and zSeries► A native Unix like operating system for IBM System z and zSeries

► “zLinux” is already trademarked, so we can’t call it that…

� z/VSE► Small- to mid-size 390 customers

► Most often used on very small 390 hardware

� z/TPF► Primarily for very high transaction rate, specialized applications (airline scheduling)

Operating System

14


Mixed workload - Virtualization - Consolidationeven beyond a single footprint, when using clustering

JavaAppl.

WebSphere

Business Logic,

Web Serving, ...

Legacy

Database &

Transaction

Processing

Native Linux

DB, Business Logic

Web Serving, ...ERP

CRM

IBM System z9

MQ DB2IMS CICS

IMS

JavaAppl.Business

Objects

z/OS z/VM

LinuxLinux

z/OS

CICS

DB2

IMS

SAP

Unix

System

services JVM

Business Objects

z/VM

Java Appl

Java Appl

Linux

for

zSeries

c++

JavaDB

Linux for

zSeriesLinux

for

zSeries

PR/SM up to 60 partitions 100’s and 100’s !! of virtual linux servers – “virtual blades”

Linux

Virtual

Servers

HiperSocketsfor integration

Unix System Services

Responsive - Variable - Resilient – Focused Autonomic - Virtualized - Open - Integrated


Summary

� From System/360 in 1964 to today’s System z9 and zS eries, we have seen an evolution that has preserved customer inves tments in a unique way

� From OS/360 to MVS to OS/390 to z/OS, we have seen an e volution of the operating system that is core to most corporate IT environments

� From CP/67 as a research project and VM/370 as a migr ation tool, VM has evolved to today’s z/VM as the core of IBM’s zSe ries virtualization technology

� Virtualization is now considered “standard” in the in dustry and all virtualization solutions owe much to the VM family

“Legacy systems are systems that work!”

15


2 system z ibm system z hardware

Documents