Nuevas Tecnologías - Arquitecturas-Avanzadas-CAECE [PDF]

Nuevas Tecnologías Si bien hay muchos fabricantes , tan solo 2 marcas en el mercado son las más conocidas y se alzan como competidoras muy serias. Hace algunos años Intel estaba a la cabeza tanto de fabricación como de ventas y distribución, pero un cambio en el mercado con una fuerte aparición de AMD ha establecido un nuevo estándar en las configuraciones de los equipos de venta directa.

Porqué retrasar el reemplazo de PCs no es una buena decisión tecnológica ni financiera?

“El consejo que continuamente les doy a mis colegas de IT es que demuestren el valor que trae al negocio cualquier gasto en tecnología. Habiendo sido CIO, siempre he creído que los gastos en tecnología deben verse como un activo para la compañía, y no como un costo para mantener operando el negocio” Stacy Smith, Intel CFO

El dilema actual • Derivado de la crisis económica mundial muchas organizaciones están retrazando el reemplazo de equipo de cómputo para reducir sus costos

• Pareciera una decisión correcta (menos dinero que gastar), sin embargo. . . – Qué hay del costo asociado de mantener equipo viejo ? (mayor a 3 años) – Qué hay del mayor consumo de energía asociado a equipos viejos ? – Qué hay de los mayores riesgos de seguridad ? – Qué hay del costo por m² que ocupan los servidores viejos ?

Porqué los equipos a partir de 3 años cuestan más? •

En promedio el costo de mantenimiento de un equipo después del 3er año se incrementa en un 59% (*) Impresoras (6%)

Equipo LAN (7%)

•

Un equipo de 3 años o más consume 80% más energía eléctrica (*) – Los datacenters acaparan la atención debido a la concentración de equipos, sin embargo las PC‟s y sus periféricos consumen mucha más energía

Equipo Telecoms móviles (9%) Equipo Telecoms Fijas (15%)

PCs & Monitores (39%)

Servidores incluyendo enfriamiento (23%)

•

Equipos de 4 años de antigüedad pueden experimentar un 53% de incremento en incidentes de seguridad

(*)

(*) “Using Total Cost of Ownership to Determine Optimal PC Refresh Lifecycles”, Wipro Technologies, March 2009 (www.wipro.com/industryresearch). Fuente: Gartner Inc. “Tera-Architectures A Convergence of New Technologies” by Martin Reynolds July26, 2007

Intel sugiere considerar los siguientes aspectos al evaluar una renovación de equipo de cómputo • Productividad de los empleados • Costos energía eléctrica & eficiencia energética • Costos operativos (administración y soporte) • Costos asociados a problemas de seguridad informática • Costos asociados al paro de labores originado por desastres naturales o problemas de salud pública

6

Intel® vPro™ La tecnología Intel® vPro™ ayuda a reducir los costos de mantenimiento de PCs mediante • Desempeño inigualable • Menor consumo de energía • Diagnóstico y reparación remota • Administración de Activos fuera de línea • Aislamiento y recuperación de PCs/Laptops infectadas

E8500

Multitarea con aplicaciones de productividad1

Hojas de cálculo complejas + virus scan2

Consumo de energía promedio4

(mayor es mejor)

(menor es mejor)

(menor es mejor)

3.6x más rápido

6.7x más rápido

1.5x mejor

55 segundos

66.0 watts

6min. 13 seg.

100.9 watts

202

P4 630

(WinMark)

54.8

(WinMark)

Intel® vPro™ La tecnología Intel® vPro™ ayuda a reducir los costos de mantenimiento de PCs mediante • Toda la funcionalidad de una PC más los beneficios de la movilidad • Mismas capacidades de administración de activos • Diagnóstico y reparación remota • Administración de Activos fuera de línea • Aislamiento y recuperación de PCs/Laptops infectadas

• Menor consumo de energía

P8600

T2700

Hojas de cálculo complejas + virus scan2

Word + PowerPoint + virus scan2

Consumo de energía promedio4

(menor es mejor)

(menor es mejor)

(menor es mejor)

1.2x más rápido

1.3x más rápido

1.6x mejor 1hr + duración de batería

1 min. 45 seg

1 min. 50 seg

20.3 watts

2 min. 8 seg

2 min. 24 seg.

33.0 watts

Intel® Xeon® Comparado con servidores de hace 4 años • Hasta 9 veces mayor desempeño por servidor • Con un 18% de reducción en consumo de energía

• Capacidad de consolidación 9 a 1 • Hasta 90% menor costo de operación • 89% reducción de espacio físico • 92% reducción consumo de energía (anual) • Retorno de inversión estimado de 8 meses

En 2005. . .

HOY

9:1 184 servidores Xeon® de 1 núcleo

21 servidores Xeon® 5500 de 4 núcleos 90% en reducción de costos de energía

Arquitecturas Intel® de Servidor para la Empresa

Intel proporciona elección de arquitectura y flexibilidad en la evolución a los 64 bits Arquitectura actual

Beneficios

Rendimiento proporciona líder, libre deelección 64 de bits ej. Database, ERP, BI,soluciones HPC OS y desde el centro de fabricante datos hasta el puesto cliente

Arquitectura elegida

Intel Arquitectura RISC

arquitectura

IA-32

Memoria Extendida cuando se necesite. Gran rendimiento en 32 bits

Rendimiento, robustez y escalabilidad líderes. Arquitectura 64bits pura

Precio/prestaciones líderes con posibilidad de direccionamiento 64 bits

* Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate performance of Intel products as measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance.

Arquitectura Intel® Itanium 2

¿Por qué Itanium?: Más que un procesador de 64 bits EPIC=Proceso de instrucciones explícitamente paralelo

EPIC 2 años

Rendimiento

SuperScalar + 9 años

RISC CISC

+ 20 años

10-15 años

 Nueva

arquitectura  Paralelismo explícito  Predicación  Especulación  Recursos masivos  Compatibilidad IA-32  Y por supuesto, direccionamiento de 64 bits

Tiempo

La arquitectura para los próximos 25 años

Arquitecturas Tradicionales:

Paralelismo Limitado Código fuente original

Código máquina secuencial

Hardware Código paralelizado

Compilación

Multiples unidades functionales

Unidades de ejecución disponibles usadas ineficientemente

.. .

.. .

.. .

Los procesadores actuales están con frecuencia hasta un 60% desocupados

.. .

Arquitectura de Microprocesador Itanium™: Paralelismo Explicito Código fuente original

Código máquina Ya paralelizado Compilación

Compilador

Múltiples unidades funcionales

Hardware Un compilador optimizado para Itanium™ analiza el código en su totalidad al compilar

Uso más eficiente de los recursos de ejecución

.. .

.. .

.. .

.. .

Optimiza la ejecución en paralelo

Procesador Intel® Itanium® 2 9M Procesadores MP, máximo rendimiento, 130W • Itanium® 2 Processor 1.60GHz, 9MB • Itanium® 2 Processor 1.6GHz, 6M • Itanium® 2 processor 1.5GHz, 4M

Procesadores DP, líderes en $/FLOP, solo DP • Itanium® 2 Processor 1.60GHz, 3MB, 400MHz FSB, 99W • Itanium® 2 Processor 1.6GHz, 3M, 533MHz FSB, 99W Procesador de bajo consumo • Itanium® 2 processor 1.3GHz, 3M, 62W

Mayor rendimiento, compatibilidad hardware

Roadmap de procesadores Intel® Itanium Plataformas Multi-procesador (MP)

Plataformas Intel® Itanium® 2 MP

Common Platform Arch. Tukwila

Montecito/Montvale

(Madison 9M) Plataformas de dos procesadores (DP)

Plataformas Intel® Itanium® 2 DP e Itanium 2 LV (Fanwood)

   

Arquitectura EPIC MCA mejorado Power Mgmt IPMI 2.0

• • • • • • •

Dual-core Multi-threading Tecnologías Foxton, Pellston Power Mgmt (Pconfig /PSMI) Tecnologías Silvervale Fully Buffered DIMMs Gestionabilidad: Redirección de consola (KVM, IDE-R),

Futuros chipset

• • • • •

Multi-core Virtualizacion mejorada Mejoras en I/O & memoria Mejoras RAS Common platform architecture • Gestionabilidad : Auto Provision, Auto Protección

+

+ Plataformas actuales

Futuras Plataformas

Millington (+ LV)/ DP Montvale (+ LV)

2005-2006+

Futuro

Dual core y después: Multi-Core Hoy Single Core

2005-2006 Dual Core

Futuro Multi-Core 4 o más cores

Cache Core

Cache

Core Cache

Core + Cache

• Futuras tecnologías de fabricación de Intel (90nm, 65nm) permitirán multiples cores manteniendo el tamaño de “die” y el consumo dentro de los límites actuales Dual Core es sólo el principio… All products, dates and features are preliminary and subject to change without notice

Procesadores Intel Xeon e Intel Xeon MP

¿Qué es Intel® EM64T? Funcionalidades

Modos

Direccionamiento de memoria extendido

Legacy Mode 32 OS / 32 Apps

64-bit Pointers, 64-bit Registers

+

Registros Adicionales 8-SSE & 8-Gen Purpose Calculo Enteros de Doble Precision (64-bit) Soporte de Espacio de direcciones plano

+

Compatibility Mode 64 OS / 32 Apps 64-bit Mode 64-bit 64/64 64 OS / 64 Apps

= Con Intel® EM64T

Evolucion de la arquitectura IA-32 con mejoras para soportar extensiones de memoria a partir de mediados de 2004

EM64T Modos soportados EM64T Modo Legacy

Modo Compatibilidad

Modo 64-Bit

• SO 32-bit OS

• SO 64-bit

• SO 64-bit

• Aplicaciones 32bit

• Aplicaciones 32-bit

• Aplicaciones 64-bit

• Drivers 64-bit

• Drivers 64-bit

• Espacio de direccionamiento 4 GB

• Espacio virtual de direcciones plano de 64-bit

• Drivers 32-bit

• Registros (GPR) 32-bit

• Registros (GPR) 64-bit

Procesador Intel® Xeon™: La nueva generación de plataformas empresariales hasta 2 procesadores Procesador Intel® Xeon™ a 3.60 GHz con FSB 800 MHz y 2 MB de Anunciado el 14 caché de nivel 2 de Febrero Intel® E7525 Chipset

Chipset Intel® E7520 Chipset Intel® E7320 Procesador Intel® IOP332 I/O Arquitecturas Optimizadas para Servidor y Workstation

Procesador Intel® Xeon™ MP: 2 nuevas plataformas multiprocesadoras • Plataforma MP de Rendimiento – Procesador Intel® Xeon™ MP 64-bit • 3.33 GHz / 8MB L3 cache • 3.00 GHz / 8MB L3 cache • 2.83 GHz / 4MB L3 cache

– Chipset Intel® E8500

• Plataforma MP de Valor – Procesador Intel® Xeon™ MP 64-bit • 3.66 GHz / 1MB L2 cache • 3.16 GHz / 1MB L2 cache

– Intel® E8500 Chipset

Anunciados el 29 de Marzo

Nueva plataformas Intel® Xeon™ RAS Reliability Availability Serviceability

bus con corrección errores ECC, mirroring & RAID de memoria (solo MP)

Consumo de Potencia

New

New

Demand Based Switching

XD bit New

RAID Inteligente

New

Nueva Generación de BUS E/S: PCI Express

Memoria más rápida: DDR2

New

New

nuevos chipsets avanzados

Extensiones 64-bit

Bus de Sistema más rápido: 800MHz (DP) y doble bus a 667MHZ (MP) New

Mejoras en la Arquitectura: SSE3, HT optimizado

Capacidad de crecimiento

Procesador Xeon a 3.6GHz y 2 MB L2 cache Procesador Xeon MP a 3.33GHz y 8 MB L3 cache

Rendimiento

Mirando hacia el futuro • Nuevas Tecnologías que permitan el crecimiento futuro de su negocio – – – – –

Soporte 64-Bit Multi-Core Tecnología de Virtualizacion Intel® Tecnología Intel® de Aceleración de E/S Tecnología Intel® Active Management

Núcleos múltiples : Rendimiento a través del Paralelismo

Rendimiento Normalizado vs. el primer procesador Intel® Pentium® 4

Performance

MULTI-CORE

10X

Estamos aquí

SINGLE CORE

3X 2004

2000

Previsión de futuro Source: Intel

2008+

Resumen • Una sola plataforma de servidor no responde a todas las necesidades de la informática empresarial actual

• La gama de procesadores para servidores de Intel es la más completa y se adapta a cada tipo de necesidad • La migración progresiva a los entornos de 64 bits permitirá a los usuarios de bases de datos eliminar cuellos de botella y obtener mayor rendimiento • Futuras tecnologías como los núcleos múltiples o la tecnología de Virtualización acentuarán aún más el liderazgo de los procesadores Intel para servidores

Recursos de referencia Estimador para Renovación de PCs • Estimador de TCO www.intel.com/business/business-pc/roi/demo.htm

Estimador para Renovación de Servidores • Estimador de TCO www.intel.com/go/xeonestimator Porqué Renovar con Tecnología Intel® vPro™ • Estimador de ROI en línea para calcular los ahorros con vPro™ www.intel.com/business/business-pc/roi/demo.htm

AMD Habla De Su Primera Arquitectura Completamente Nueva • Nuestro equipo editorial, ha llevado a cabo una excelente labor para informales todo lo último que ha "comentado" AMD durante el AMD Analyst Day en relación a su futuro en el sector de los procesadores. Así, ya les hemos contado sobre sus futuras plataformas deEscritorio, Móviles y Servidores, les hemos hablado del Sandtiger Octa-Core, de la tecnología G3 Memory Extender, los procesos de fabricación, y por supuesto del ya conocido Barcelona. Sin embargo, esto no es todo, y a continuación queremos introducirlos en algunos nuevos conceptos, fundamentales para AMD, que son la base de todo lo recién indicado y que permitirá complementar la excelente cobertura ya entregada y mencionada en las líneas superiores.

1.- M-Space. • Tal como se indicó, AMD planea lanzar su plataforma Opteron de tercera generación, el año 2009, con el procesador Sandtiger OctaCore. Bajo este procesador, se encuentra un nuevo acercamiento de la Compañía, modular, denominado M-Space. De acuerdo a la gente de AMD, M-Space permite mezclar y calzar funciones de una CPU para determinadas tareas.

"Acercamiento modular de M_Space"

Definición de M-SPACE •

La definición de M-Space, es la siguiente, – a.- Modular: "Bloques de diseño" reconfigurables para mayor agilidad y velocidad. – b.- Escalable: Escalamiento linear de rendimiento "multi-thread" y "single-thread“. – c.- Portabilidad: Eficiencia energética para mayor portabilidad y movilidad. – d.- Accesibilidad: Compromiso a una innovación abierta. – e.- Compatible: Compatibilidad hacia atrás y facilidad de actualización. – f.- Eficiencia: Eficiencia óptima en el chip y a nivel de Entrada/Salida.

"Definición del concepto M-Space"

• Como se puede apreciar, M-Space es una completa redefinición de la estrategia de diseño de AMD, la cual, como se mencionó, es la base del Sandtiger. • Pese a lo anterior, este nuevo procesador también incluye otro concepto, no menos interesante, eje central de los planes futuros de la Compañía, el cual pasamos a revisar a continuación.

2.- Bulldozer. • El Sandtiger de 8 núcleos de AMD, consiste en 8 Bulldozers. Suena extraño el nombre, el cual ha sido asociado a una intención, por parte de la Compañía, de demoler el mercado con este producto. • Bulldozer, es el nombre que la Empresa le ha dado a uno de los núcleos de CPU que se basan en la ya mencionada arquitectura M-Space. AMD ha indicado que se obtendrán mejoras dramáticas en rendimiento por watt, en aplicaciones de Computación de Alto Rendimiento ("High Performance Computing" o "HPC"), al usar los núcleos Bulldozer.

"Objetivo del nuevo núcleo Bulldozer"

• Lo interesante del tema, es que a diferencia de Barcelona y Shangai, los cuales corresponden a una evolución de la longeva arquitectura K8, Bulldozer es algo totalmente nuevo y diferente. • AMD colocará 8 núcleos Bulldozer en el Sandtiger, con un controlador de memoria. Segúnla Compañía, el concepto detrás de M-Space podrá verse en este producto al optimizarse el diseño para Servidores y elevar la vara del rendimiento por watt, tanto para aplicaciones "single-threaded" como para aquellas "multi-threaded". Algunos expertos ya hablan de plataformas capaces de "mutar" de acuerdo a los requerimientos de las aplicaciones.

3.- Fusion. • La arquitectura M-Space, también puede relacionarse con la comentada tecnologíaFusion, la cual mezcla CPUs y GPUs. • AMD planea mezclar y calzar componentes del concepto M-Space en Falcon, un procesador Fusion optimizado para Computadores de Escritorio mainstream, y paraPortátiles. Falcon es la base de la plataforma de rango medio de Escritorio de AMD, llamada Copperhead.

"Componentes del Falcon-Fusion"

• Falcon cuenta con cuatro núcleos Bulldozer, junto a un procesador gráfico integrado. Esta GPU, soporta tecnología DirectX 10, posiblemente DX11, y cuenta con la tecnología de AMD denominada Universal Video Decoder o UVD. También posee un PCIe integrado.

4.- Bobcat. • Junto con Bulldozer, AMD cuenta con el núcleo Bobcat, también para procesadoresFusion, diseñado para aplicaciones moviles, ultra-móviles y de electrónica de consumo.Bobcat también es un diseño completamente nuevo y cuenta con grandes capacidades de escalamiento energético. Los procesadores basados en Bobcat pueden llegar a consumir sólo 1 watt de energía. No se conocen mayores detalles de los procesadoresFusion con "tecnología Bobcat".

"El Bobcat dirigido al sector móvil"

5.- GPUs. • Cuando AMD compró ATI, obviamente era con una idea fija en mente. En efecto, la Compañía está plenamente convencida que el desarrollo gráfico, y por supuesto la ya tan comentada integración con CPUs, es más importante que dedicarse sólo al diseño deprocesadores. Obviamente, ambos conceptos están fuertemente arraigados e interrelacionados en los Roadmaps de la Compañía.

"Nuevas tecnologías gráficas de AMD-ATI, primera parte"

• Después del desastre que se ha obtenido con la incursión de AMD en el mundo DirectX 10, muchos pueden mirar con cierta mofa a la Compañía, cuando ésta se refiere a gráficos de la siguiente generación.

"Nuevas tecnologías gráficas de AMD-ATI, segunda parte"

• Para el común de las personas, ATI es sinónimo de tarjetas de video para jugar. Sin embargo, no debe olvidarse que ésta, siempre ha sido un fuerte suministrador de productos para dispositivos móviles, y AMD intenta continuar en esa línea, fortaleciéndola, al integrar tecnologías que se observan en los PCs, a los más variados dispositivosportátiles.

6.- Palabras Finales. • Tiempo atrás, AMD fue muy criticada por guardar silencio en relación a sus nuevos proyectos y tecnologías. Incluso, se llegó a decir que eso había influido en los últimos "fracasos" de la Compañía. En esa oportunidad, AMD se comprometió a ser "más abierta". • Claramente, la Empresa ha cumplido con lo prometido, pero se ha ido al otro extremo. Nos ha llenado de anuncios, conceptos, nombres, etc. Todo bien, claro que esto debe plasmarse en lanzamientos reales ya que muchos simplemente, cuando miran hacia AMD, piensan, • "…ver para creer.“ • Ciertamente, Bulldozer, sin desmerecer las otras tecnologías, será el "caballo de batalla" de AMD. Es un alejamiento definitivo de la fiel plataforma K8, pero al mismo tiempo es un tremendo desafío.

"Visión modular del futuro por parte de AMD"

SUN Microsystems Scalable Processor Architecture (SPARC)

Overview •Designed to optimize compilers and pipelined hardware implementations. • Offers fast execution rates. • Engineered at Sun Microsystems in 1985 Based on RISC I & II which were developed at University of Cal at Berkeley. • SPARC “register window” architecture

Features Performance and Economy Simplified instruction set Higher number of instructions with fewer transistors

Scalability Flexible integration of cache, memory and FPUs

Open Architecture Compatible technology to multiple vendors Now allow access to CPU component techniques Complete set of development tool available for h/w & s/w

Registers General purpose/ working data registers IU‟s „r‟ registers FPU‟s „f‟ registers

Control status registers IU control/status registers FPU control/status registers Coprocessor (CP) control/status registers

Registers Window Overlapping Each window shares its ins and outs with two adjacent windows Incremented by a RESTORE instruction decremented by a SAVE instruction Due to windowing the number available to software is 1 less than number implemented When a register is full the outs of the newest window are the ins of the oldest, which still contain valid program data

IU Control/Status Registers Processor State Register (PSR) Window Invalid Mask (WIM) Multiply/Divide (Y) Program Counters (PC, nPC) Ancillary State Registers (ASR) Deferred-Trap Queue Trap Base Register (TBR)

IU Control/Status Registers Processor State Register (PSR) Contains various fields that control and hold status information Impl 31:28

Ver 27:24

Icc 23:20

Reserved 19:14

EC 13

EF 12

PIL 11:8

S 7

PS

ET

6

5

CWP

Window Invalid Mask (WIM) To determine a window overflow or underflow

W31 W30 W29 -----------------------

W1 W0

4:0

Memory Each location identified by Address Space Identifier (ASI) 64-bit address

Real memory No side effects

I/O locations Side effects

Snoop

Instruction Formats VIS – Visual Instruction Set Visualization built into chip

Examples of formats

Processor Comparison Summary Architecture Open versus proprietary 64-bit architecture High volume processor High bandwidth

UltraSPARC-IIi SPARC V-9 Open Ð Ð Ð

HP PA-8000 HP-PA Proprietary

Integration Level On-chip MMU On-chip I/O interface On-chip cache On-chip multimedia support

Ð Ð Ð Ð

Ð

300 MHz

Features Clock speed Binary compatibility with existing applications Performance SPECint95/fp95

Target Environment

Cyrix MediaGX X86 Proprietary Ð Ð

MIPS R10000 MIPS III Open Ð

Intel Pentium II PowerPC 603e X86 PowerPC Proprietary Open Ð Ð Ð

Ð Ð

Ð Ð Ð

Ð

180 MHz

180 Mhz

Ð

Ð

>12/>12

11.8/18.7

Low cost desktops and servers

Workstations

Low-power,

and

low-cost

servers

desktops and

Ð Ð Ð

Ð

195 MHz

233-300 Mhz

300 MHz

Ð

Ð

Ð

Ð

N/A

10.7/19.0

11.7/8.15

7.4/6.1

High-end embedded applications: networking,

Desktops

Workstations

Workstations and portables servers

What makes the CISC lock-up? Elegant forward looking branch instruction set Compiler can go to different branches

More complete testing of SPARC Simpler compiler design Better integration of OS interrupts to H/W interrupts Solaris has a tighter source code Less devices to support

References Weaver, David/Tom Germond. SPARC Architecture Manual: Version 9, Prentice Hall. 1994. Stallings, William. Computer Organization and Architecture: 5th Edition, Prentice Hall. 2000. Bresani, Fred. Systems Engineer, Sun Microsystems. http://www.sun.com http://www.sparc.com http://www.fujitsu.com

IBM POWER 7 Series

May 2010

•POWER7 Processor

Agenda

•POWER7 Servers

 POWER 750  POWER 755  POWER 770  POWER 780

 POWER Blades •Performance/Competition •Active Memory Expansion

•Related Announcements •Upgrades •Q&A

Customers are Moving to Higher Value …as shown by the largest shift of customer spending in UNIX History

UNIX Server Rolling Four Quarter Average Revenue Share POWER6 POWER5™

Micro-Partitioning

Live Partition Mobility

POWER4™

Dynamic LPARs

Source: IDC Quarterly Server Tracker Q309 release, November 2009

POWER7 System Highlights •Balance System Design – Cache, Memory, and IO

•POWER7 Processor Technology – 6th Implementation of multi-core design – On chip L2 & L3 caches •POWER7 System Architecture – Blades to High End offerings – Enhances memory implementation – PCIe, SAS / SATA •Built in Virtualization – Memory Expansion – VM Control

•Green Technologies – Processor Nap & Sleep Mode – Memory Power Down support – Aggressive Power Save / Capping Modes

600

500

400

300

•Availability – Processor Instruction Retry – Alternate Process Recovery – Concurrent Add & Services

200

100

0

JS23

JS43

520

550

560

570/16

570/32

595

73

Power your Planet

+

AIX - the future of UNIX Total integration with i Scalable Linux ready for x86 consolidation

Workload-Optimizing Systems Virtualization without Limits  Drive over 90% utilization

Dynamic Energy Optimization  70-90% energy cost reduction

 Dynamically scale per demand

 EnergyScale™ technologies

Resiliency without Downtime  Roadmap to continuous

Management with Automation  VMControl to manage

availability  High availability systems & scaling

virtualization  Automation to reduce task time

Smarter Systems for a Smarter Planet.

POWER7 Processor POWER7 Processor

IBM POWER Processor Roadmap 3 Year Revolution 18 month “+” evolution POWER8 POWER7/7+

POWER6/6+ POWER5/5+ POWER4/4+

Hardware Virtualization First Dual Core for Unix & Linux Dual Core & Quad Core Md in Industry Enhanced Scaling  Dual Core  Chip Multi Processing  Distributed Switch  Shared L2  Dynamic LPARs (32) 180nm,

2001

2 Thread SMT Distributed Switch + Core Parallelism + FP Performance + Memory bandwidth + 130nm, 90nm

2004

Fastest Processor In Industry          

Dual Core High Frequencies Virtualization + Memory Subsystem + Altivec Instruction Retry Dyn Energy Mgmt 2 Thread SMT + Protection Keys 65nm

2007

Most POWERful & Scalable Processor in Industry         

IBM is the leader in Processor and Server design

4,6,8 Core 32MB On-Chip eDRAM Power Optimized Cores Mem Subsystem ++ 4 Thread SMT++ Reliability + VSM & VSX Protection Keys+ 45nm, 32nm

2010

Future

Cores:        

Transition from POWER6 POWER7

8 Intelligent Cores / chip (socket) 4 and 6 Intelligent Cores available on some models 12 execution units per core Out of order execution 4 Way SMT per core 32 threads per chip L1 – 32 KB I Cache / 32 KB D Cache per core L2 – 256 KB per core

Chip:  32MB Intelligent L3 Cache on chip

Memory:  Dual DDR3 Controllers  100 GB/s sustained Memory bandwidth / chip

Scalability:    

Up to 32 Sockets 360 GB/s peak SMP bandwidth / chip 590 GB/s peak I/O bandwidth / chip Up to 20,000 coherent operations in flight

P Core Core O W L2 L2 E R G L3 Cache X eDRAM B U L2 S Core

Core

L2

L2

L2

L2

L2

Core

Core

Core

Memory Interface

Energy:  Aggressive processor Nap & Sleep modes  10% “Over clock” when thermals are good

Core

Memory++

S M P F A B R I C

Memory Channel Bandwidth Evolution POWER5

POWER6

POWER7

Memory Performance: 2x DIMM

Memory Performance: 4x DIMM

Memory Performance: 6x DIMM

D D R 3

DDR2 @ 553 MHz Effective Bandwidth: 1.1 GB/s

DDR2 @ 553 / 667 MHz Effective Bandwidth: 2.6 GB/sec

D D R 3

D D R 3

D D R 3

D D R 3

D D R 3

D D R 3

D D R 3

D D R 3

DDR3 @ 1066 MHz Effective Bandwidth: 6.4 GB/sec

D D R 3

Multi-Threading Evolution Single thread Out of Order

S80 HW Multi-thread FX0 FX1 FP0 FP1 LS0 LS1 BR C XR L

FX0 FX1 FP0 FP1 LS0 LS1 BR CR X L

POWER5 2 Way SMT

POWER7 4 Way SMT

FX0 FX1 FP0 FP1 LS0 LS1 BR CR X L

FX0 FX1 FP0 FP1 LS0 LS1 BR CR X L No Thread Executing

Thread 0 Executing Thread 2 Executing

81

Thread 1 Executing Thread 3 Executing

POWER7 TurboCore Mode Power 780 POWER7 TurboCore Chip Chip

 TurboCore Chips: 4 available cores  Aggregation of L3 Caches of unused cores.  TurboCore chips have a 2X the L3 Cache per

Chip available

–4 TurboCore Chips

L3

= 32 MB  Performance gain over POWER6.

–Provides up to 1.5X per core

to core  Chips run at higher frequency:

–Power reduction of unused cores. TurboCores  With “Reboot”, System can be reconfigured to 8 core mode. Unused –ASM Menus Core

P O W E R

Core Core Core Core L2

G X B U S

L2

L2

L2

32 MB L3 Cache

L2

L2

L2

L2

Core Core Core Core Memory Interface

S M P F A B R I C

POWER7 Multi-Threading Options •TurboCore Option •50% of the available cores active

Standard Option All cores active

3

3

2.5

2.5

2

2

1.5

1.5

1

1

0.5

0.5

0

0

SMT4

SMT2

Single

SMT4

SMT2

Single

POWER7 Servers

Power Systems Portfolio Power 780

Consistency    

Binary compatibility Mainframe-inspired reliability Advanced Virtualization AIX, Linux and IBM i OS

Complete flexibility for workload deployment

Power 595

Power 770 Power 570 Power 750 Power 550

Power 700 701 & 702 JS Blades

Power 520

Power 755

HPC

Power 575

Power is the Innovation that will Deliver Business Advantages Leadership Performance and Energy Efficiency

 Industry leading performance with POWER7 processors Power 750 Express: 1- to 4-socket; 6- or 8-cores per socket

 ENERGY STAR-qualified Meets EPA guidelines for energy efficiency

 Workload-optimizing capabilities improve performance

Power 750 Express

Intelligent Threads optimization, Intelligent Cache sharing, Active Memory Expansion and 320 virtual machines in 2010

 Intelligent Energy capabilities that balances performance and efficiency: Frequency boost for increased performance or reduction during low demand for energy reduction >3X increased performance per watt

 Ease of Ownership Popular, ready to run „Edition‟ configurations Light Path Diagnostics Customer Setup

Power 750 System 8233-E8B POWER7 Architecture DDR3 Memory

System Unit SAS SFF Bays

4U Depth: 28.8”

System Unit IO Expansion Slots

Integrated SAS / SATA System Unit Integrated Ports Integrated Virtual Ethernet System Unit Media Bays IO Drawers w/ PCI slots Cluster Redundant Power and Cooling Certification (SoD) EnergyScale

6 Cores @ 3.3 GHz 8 Cores @ 3. 0, 3.3, 3.55 GHz Max: 4 Sockets Up to 512 GB Up to 8 Drives (HDD or SSD) 73 / 146 / 300GB @ 15k (2.4 TB) (Opt: cache & RAID-5/6) PCIe x8: 3 Slots (2 shared) PCI-X DDR: 2 Slots 1 GX+ & Opt 1 GX++ 12X cards Yes

3 USB, 2 Serial, 2 HMC Quad 10/100/1000 Optional: Dual 10 Gb 1 Slim-line DVD & 1 Half Height PCIe = 4 Max: PCI-X = Max 8 12X SDR / DDR (IB technology) Yes (AC or DC Power) Single phase 240 VAC or -48 VDC NEBS / ETSI for harsh environments Active Thermal Power Management Dynamic Energy Save & Capping

750 CPW & rPerf Details 6-core

8-core

8-core

3.3 GHz

CPW

rPerf

6-core

37200

70.07

12-core

69200

134.54

18-core

94900

193.40

24-core

135300

252.26

8-core

44600

81.24

16-core

82600

155.99

24-core

122500

224.23

32-core

158300

292.47

8-core

47800

86.99

16-core

88700

167.01

24-core

129700

140.08

32-core

168800

313.15

181000

331.06

3.0 GHz

3.3 GHz

#8335

#8334

#8332

8-core 3.55 GHz #8336 32-core

POWER6

rPerf

550 (8) 4.7Ghz

68.20

550 (8) 5.0GHz

78.60

560 (16) 3.6GHz

100.3 0

570 (16) 4.4GHz

127.3 2

570 (16) 5.0GHz

141.2 1

595 (32) 4.2GHz

266.5 1

595 (64) 4.2GHz

479.8 9

595 (32) 5.0GHz

307.1 2

595 (64) 5.0GHz

553.0 1

Power 750 vs Power 550 / 560 Performance* / K BTU

Performance* / KW 180

70

160

60

140

50

120 100

40

80

30

60

20

40 20

> 406% Improvement

0

10

> 421% Improvement 0

Power 750

Power 550

Power 560

Power 750

Power 550

Power 560

Power 755 4-Socket HPC System Power 755

4U x 28.8” depth Up to 8.4 TFlops per Rack ( 10 nodes per Rack )

5.3 / 6.1 RHEL / SLES

POWER7 Architecture

4 Processor Sockets = 32 Cores 8 Core @ 3.3 GHz

DDR3 Memory

128 GB / 256 GB, 32 DIMM Slots

System Unit SAS SFF Bays

Up to 8 disk or SSD 73 / 146 / 300GB @ 15K (up to 2.4TB)

System Unit Expansion Integrated Ports

PCIe x8: 3 Slots (1 shared) PCI-X DDR: 2 Slots GX++ Bus 3 USB, 2 Serial, 2 HMC

Integrated Ethernet

Quad 1Gb Copper (Opt: Dual 10Gb Copper or Fiber)

System Unit Media Bay

1 DVD-RAM ( No supported tape bay )

Cluster

Up to 64 nodes Ethernet or IB-DDR

Redundant Power

Yes (AC or DC Power) Single phase 240vac or -48 VDC

Certifications (SoD)

NEBS / ETSI for harsh environments

EnergyScale

Active Thermal Power Management Dynamic Energy Save & Capping

Power 755 vs. 750 Offering Structures Feature

750

755

Processors

32-core @ 3.3 GHz

8 / 16 / 24 / 32-core @ 3.55 GHz 6 / 12 / 18 / 24-core @ 3.3 GHz 8 / 16 / 24 / 32-core @ 3.0 GHz

Memory

128GB - 256GB 4GB & 8GB DIMMS

8 - 512GB 4GB, 8GB, 16GB DIMMS

GX slot support

Yes – IB clustering

Yes

I/O Drawer support

No

Yes

DASD Backplane

No Split Backplane

Split Backplane support

Integrated Ethernet

Quad GbE or Dual 10GbE

Quad GbE or Dual 10GbE

Virtualization

No PowerVM support

PowerVM Std and Ent

System unit SFF Bays and drives

8 SFF SAS HDD / SDD 10k and 15K SFF drives

8 SFF SAS HDD / SDD 10k and 15K SFF drives Optional Integrated RAID

Internal Tape

No

Yes

Performance Metric

TFLOPS

rPerf

Operating System

AIX, Linux No H/W Raid Cards

AIX, IBM i, Linux H/W Raid Cards

93

Power 770 Power 770 Processor Technology

4U x 32 inches Depth

Maint. Coverage: 9 x 5

L3 Cache Redundant Power & Cooling Redundant Server Processor Redundant Clock Concurrent Add Support Concurrent Service System Unit Processors DDR3 Memory (Buffered) SAS/SSD SFF Bays DVD-RAM Media Bays SAS / SATA Controller PCIe bays GX++ Slots (12X DDR) Integrated Ethernet USB 12X I/O Drawers w/ PCI slots

6 Cores @ 3.55 GHz 8 Cores @ 3.1 GHz On Chip Yes Yes / Two Enclosure minimum

Yes / Two Enclosure minimum Yes Yes Single Enclosure 4 Enclosures Up to 2 Sockets 8 Sockets Up to 512 GB Up to 2 TB 6 24 1 Slim-line 4 Slim-line 2/1 8/4 6 PCIe 24 PCIe 2 8 Std: Quad 1Gb Std: Four Quad 1Gb Opt: Dual 10Gb + Opt: Four x Dual 10Gb Dual 1 Gb + Dual 1 Gb 3 12 Max: 4 PCIe, 8 Max: 16 PCIe, 32 PCI-X PCI-X 94

Power 780 Power 780 Processor Technology

4 Cores @ 4.14 GHz 8 Cores @ 3.86 GHz On Chip

TurboCore

L3 Cache Redundant Power & Yes Cooling Redundant Server Yes / Two Enclosure minimum Processor Redundant Clock Yes / Two Enclosure minimum Concurrent Add Support Yes Concurrent Service Yes Single Enclosure 4 Enclosures System Unit Processors 2 Sockets 8 Sockets

Maint Coverage 24 X 7

PowerCare Support

DDR3 Memory (Buffered) SAS/SSD SFF Bays (CEC) DVD-RAM Media Bays SAS / SATA Controller PCIe (CEC) GX++ Slots (12X DDR) Integrated Ethernet USB 12X I/O Drawers w/ PCI slots

Up to 512 GB

Up to 2 TB

6

24

1 Slim-line 4 Slim-line 2/1 8/4 6 PCIe 24 PCIe 2 8 Std: Quad 1Gb Std: Four Quad 1Gb Opt: Dual 10Gb + Opt: Four x Dual 10Gb Dual 1 Gb + Dual 1 Gb 3 12 Max: 4 PCIe, 8 Max: 16 PCIe, 32 PCI-X PCI-X 95

770 and 780 CPW & rPerf Details 770

12-core

16-core

8-core

780

3.5 GHz

CPW

rPerf

12-core

73100

140.75

24-core

99000

261.19

36-core

131050

377.28

550 (8) 4.7Ghz

68.20

48-core

248550

493.37

550 (8) 5.0GHz

78.60

560 (16) 3.6GHz

100.30

570 (16) 4.4GHz

127.32

3.1 GHz

#4980

#4981

POWER6

rPerf

16-core

88800

165.30

570 (16) 5.0GHz

141.21

32-core

155850

306.74

595 (32) 4.2GHz

266.51

48-core

229800

443.06

595 (64) 4.2GHz

479.89

64-core

292700

579.39

595 (32) 5.0GHz

307.12

CPW

rPerf

595 (64) 5.0GHz

553.01

16-core

105200

195.45

32-core

177400

362.70

48-core

265200

523.89

64-core

343050

685.09

3.86 GHz

#4982

780 TurboCore mode values not shown

Power 770 & 780 vs Power 570 Differences Power 570

Power 770 & 780

Up to 8 sockets, Up to 32 Cores

Up to 8 Sockets, Up to 64 cores

Up to 768 GB Memory

Up to 2 TB Memory ( Initial GA will be 1 TB)

DDR2 DIMMS

DDR3 DIMMS

Six 3.5” SAS Bays / Enclosure

Six SFF SAS Bays / Enclosure

4 PCIe & 2 PCI-X slots per Enclosure

6 PCIe slots per Enclosure

No write cache or RAID-5/6 support

Write cache & RAID-5/6 support

Single integrated DASD / Media Cntlr

Three integrated DASD / Media Controllers

Optional Split Backplane

No Power & Management Thermal Clock Cold Failover ECC with bit steer Concurrent Drawer Maint restrictions Concurrent Drawer Add cable restrictions

Standard Split backplane Optional Tri-Split Backplane Power & Thermal management TPMD support Clock Hot Failover ECC with DRAM sparing No Restrictions ( 4Q / 2010 ) No Restrictions

97

IBM Power Systems Comparisons Nodes Cores (single system image) Frequency SMP buses System memory Memory per core Memory Bandwidth (peak)

Power 750

Power 770

Power 780

Power 595

One 6, 12, 18, 24 or 8, 16, 24, 32 3.0, 3.3, 3.55 GHz 4 byte Up to 512 GB 16 or 21 GB

Up to four

Up to four

4 – 64

4 – 64

3.1, 3.5 GHz 8 byte Up to 2 TB* 32 or 42 GB

3.8, 4.1 GHz 8 byte Up to 2 TB* 32 or 64 GB

Up to eight 8 – 64 Upgradeable to 256 4.2, 5.0 GHz 8 byte Up to 4 TB 64 GB

273 GB/s

1088 GB/s

1088 GB/s

1376 GB/s

Memory Bandwidth per core (peak) Memory controllers

8.5 GB/s

17 or 22 GB/s

17 or 34 GB/s

21.5 GB/s

1 per processor

2 per processor

2 per processor

2 per processor

I/O Bandwidth (peak)

30 GB/s

236 GB/s

236 GB/s

640 GB/s

I/O Bandwidth per core (peak) I/O loops Total disk drives rPerf per core Maximum LPARs

0.9GB/s

3.6 or 4.9 GB/s

3.6 or 7.3 GB/s

10 GB/s

Up to 2 Up to 576 Up to 11 Up to 320*

RAS

Standard

Warranty PowerCare

9x5 No

Up to 8 Up to 1200 Up to 11 Up to 640* Enhanced Memory Dynamic FSP & clocks 9x5 No

Up to 8 Up to 1200 Up to 13 Up to 640* Enhanced Memory Dynamic FSP & clocks 24 x 7 Yes

Up to 32 Up to 2640 Up to 10.8 Up to 254 Enhanced Memory Dynamic FSP & clocks 24 x 7 Yes

*

Planned availability in 4Q 2010

Standard

Move Up to Enterprise Class RAS RAS Item

Optional Not available Power 750

Power 770

Power 780

Redundant / Hot Swap Fans & Blowers Hot Swap DASD / Media / PCI Adapters Concurrent Firmware Update

Redundant / Hot Swap Power Supplies Dual disk controllers (split backplane) Processor Instruction Retry Alternate Processor Recovery Storage Keys PowerVM™/Live Partition Mobility/Live Application Mobility Redundant Service Processors

*

*

Redundant System Clocks

*

*

Hot-node Add / Cold-node Repair

*

*

Hot-node Repair / Hot-memory Add

*

*

Dynamic Service Processor and System Clock Failover

*

*

Hot-node Repair / Hot-memory Add for all nodes**

*

*

Redundant / Hot Swap Power Regulators Dynamic Processor Sparing Memory Sparing Hot GX Adapter Add and Cold Repair

POWER7 Enhanced Memory

Hot GX Adapter Repair

* Requires two or more nodes ** Planned for 4Q 2010

POWER7 systems are over twice as good as POWER6 systems! 

Twice the performance: Power 780 32-core performance per core is over twice the Power 570 32-core



Twice the scaling:



Twice the capacity:

Power 770 and 780 both offer twice the number of cores as the largest Power 570 Power 770 and 780 offer more than twice (~3 times) the throughput of the largest Power 570



Twice the memory: - Over twice the physical memory of the Power 570 - Active Memory ExpansionTM enables up to twice the effective memory compared to what is physically installed



Twice the energy efficiency: Power 770 & 780 offer over twice the performance per watt (up to 3 times) than the most efficient Power 570



Twice the cores for the same price: Buy twice the cores with the Power 770 and pay less than a comparable POWER6 based Power 570

POWER7 Blades

Power System Blades PS700/701/702 4 Cores @ 3.0 GHz PS700 8 Cores @ 3.0GHz PS701 16 Cores @ 3.0GHz PS702 64GB/128GB/256GB PS700/PS701/PS702

POWER7 Architecture DDR3 Memory Internal Disk

0-2 SFF DASD

PCIe Slots

2/2/4

Integrated SAS / SATA

Yes

System Unit Integrated Ports

3 USB, 2 Serial, 2 HMC

Integrated Virtual Ethernet

Quad Gigabit Optional: Dual 10 Gb

EnergyScale

Active Thermal Power Management Dynamic Energy Save & Capping

Description

rPerf

CPW

8406-70Y PS700 4-cores 3.0GHz

45.13

21,000

8406-71Y PS701 8-cores 3.0GHz

81.24

42,000

8406-72Y PS702 16-cores 3.0GHz

154.36

76,300

IBM BladeCenter PS700/701/702 Express Smart BladeCenter Solutions with

Power Blades • Performance and Energy Efficiency

• Smart choices to minimize complexity, improve efficiency and scale easily • Single-wide 4 or 8-core or Double-wide 16-core • Elegantly simple scalability • Intelligent Threads • Utilizes more threads when workloads benefit • Intelligent energy optimization with EnergyScale Technology • Boosts frequency for more performance • High Performance Computing acceleration (AltiVec SIMD acceleration) • Execute up to eight single-precision or double-precision floating point operations per clock cycle per core • Flexibility and Choice

• Supports AIX, i and Linux operating systems • Consolidate all three on a single platform • Supports multiple BladeCenter chassis . 105

4, 8 or 16 cores Single or Double Wide 3.0GHz POWER7 Up to 256GB of Memory

Relative Performance

New Power Blades are better than Sun CMT blades in every important performance category

5.00 4.00 3.00 2.00 1.00 0.00 Specjbb2005

Power 702

SPECint_rate2006

Sun T6340

Relative Performance Density

• 2.4x to >3.0x better performance per blade, per core, and per socket • 2.2x to >3.9x better performance per rack unit • 1.6x to 3.0x better performance per Watt

4.00 3.00 2.00 1.00 0.00 Specjbb2005

Power 702

SPECint_rate2006

Sun T6340

Relative Energy Efficiency 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Specjbb2005

Power 702

SPECint_rate2006

Sun T6340

Power Blades run faster and cost less than other UNIX blades Reinforces why Power Blades are the most popular blades for UNIX •

Comparison to HP BL860c i2 – PS702 has > 3X performance

2 socket blade performance comparison POWER7 based blades vs Itanium 9300 based blades

600 500 400

– PS702 P/P is even better

300 200 100 0 SPECint_rate2006

SPECfp_rate2006

PS702

BL660c i2

Estimated HW Price/Performance for benchmark configurations Lower is better The Intel Itanium is tottering towards death, analyst reckons C Shanti | Thu 26th Nov 2009, 10:56 am http://www.tgdaily.com/hardware-features/44828-the-intelitanium-is-tottering-towards-death-analyst-reckons

300

200 A report from senior analyst Jon Peddie suggests that Intel's 64 bit flagship microprocessor, the Itanium, is dead in the water. So is it a turkey?

100 0

SPECint_rate2006

SPECfp_rate2006

PS702 BL660c i2

See substantiation chart for sources & benchmark detail

IBM Power Blades Deliver Lower TCA than HP x86 blade solution

At equal capacity for a full BladeCenter H Chassis with 7 two socket (16-core) PS702 blades compared to a full HP C7000 Blade Chassis with 16 two socket (12-core) HP BL460c G6 blades leveraging the higher utilization and virtualization efficiency capabilities of Power Blades. 39% lower TCA with 56% less blades HP Solution

US$347,271

IBM Solution

US$213,053

See Lower TCA chart for sources and substantiation

BladeCenter PS Blade Overview Enhances most popular blades for UNIX* with 3 New Offerings! IBM BladeCenter PS700 Express

IBM BladeCenter PS701 Express

IBM BladeCenter PS702 Express

POWER7 4-Core (1 Socket x 4 Cores per blade) Single Wide

POWER7 8-core (1 Socket x 8 Cores per blade) Single Wide

POWER7 16-core (1 Socket x 8 Cores per blade) Double Wide

4GB to 64GB DDR3 (Chipkill) 4GB@1066MHz, 8GB@800MHz

4GB to 128GB DDR3 (Chipkill) 4GB@1066MHz, 8GB@800MHz

4GB to 256GB DDR3 (Chipkill) 4GB@1066MHz, 8GB@800MHz

0-2 SAS disk

0-1 SAS disk

0-2 SAS disk

1 PCI-E CIOv Expansion Card 1 PCI-E CFFh ExpansionCard

1 PCI-E CIOv Expansion Card 1 PCI-E CFFh ExpansionCard

2 PCI-E CIOv Expansion Card 2 PCI-E CFFh ExpansionCard

Integrated Features

Keyboard, Video and Mouse Dual Port 1Gb Ethernet SAS Controller USB

Keyboard, Video and Mouse Dual Port 1Gb Ethernet SAS Controller USB

Keyboard, Video and Mouse Quad Port 1Gb Ethernet SAS Controller USB

Scalability Support

Architecture

Memory DASD / Bays Expansion Card Slots

No

Yes – Factory or Customer Upgrade

Yes – Factory or Customer Upgrade

Fibre Support

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Redundant Power

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Redundant Cooling

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Yes (via BladeCenter Chassis)

Service Processor

FSP1 (IPMI, SOL)

FSP1 (IPMI, SOL)

FSP1 (IPMI, SOL)

IBM PowerVM (optional Editions)

IBM PowerVM (optional Editions)

IBM PowerVM (optional Editions)

IBM Director and CSM IBM EnergyScale Technology

IBM Director and CSM IBM EnergyScale Technology

IBM Director and CSM IBM EnergyScale Technology

AIX, i, Linux

AIX, i, Linux

AIX, i, Linux

BCE, BCH*, BCHT, BCT, BCS*

BCH*, BCHT, BCS*

BCH*, BCHT, BCS*

* In Power Systems Channel

* In Power Systems Channel

* In Power Systems Channel

Virtualization Systems Management OS Support BladeCenter Chassis Support

* IDC 4Q2009 Server Tracker RISC/Itanium blades

Performance & Virtualization versus the competition

Power Systems Virtualization for Oracle – Tier Consolidation & Virtualization – Dynamically Resizable

8 Cores

11Core Cores Virtual I/O

CUoD

Server Partition Int Virt Linux Manager

AIX V5.3

Linux

Oracle10g Oracle11g Oracle 11g

Tier Consolidation

AIX V5.3

Oracle 9i

Storage Sharing Ethernet Sharing

3 24 3 Cores CoresCores PowerVM‟s

Linux WebSphere WebSphere WebSphere WebSphere WebSphere

Manager

6 Cores

Oracle9i Oracle 10g Oracle 10g

Int Virt Linux

8 Cores

Oracle 10g

Server Partition

16 Cores PowerVM‟s

Oracle 11g

1 Core Cores Virtual I/O

Storage Sharing Ethernet Sharing

Virtual LAN

POWER Hypervisor ISV Pricing on Power 64 core system Network

Oracle EE: 38 cores WebSphere: 1920 PVUs Do not pay for VIO server or CUoD cores

Virtual Network WebSphere to Oracle works at memory speeds

Network

Live Partition Mobility On Oracle Workloads Reduce impact of planned outages, relocate workloads to enable growth, provision new technology with no disruption to service

LPAR-1 LPAR-2

LPAR-3

LPAR-4

VIOS

VIOS

Migration Controller

Migration Controller

LPAR-1 LPAR-2 LPAR-3 LPAR-4

Oracle

Oracle

Def 1 P P P P Def 2 Def 3 P P P P P P P P P P P Def 4 AIX Kernel AIX Kernel AIX Kernel AIX Kernel Hypervisor

P

P

P PP PP

P P P

PP PP PP

P P P

AIX Kernel AIX Kernel AIX Kernel AIX Kerne Hypervisor Ethernet

Partition Mobility Requires: • POWER6

• AIX 5.3 / 6.1 or Linux • All resources must be “Virtualized” •No real resources • SAN storage environment •SAN Boot, temp space, same network

SAN Boot Data

Partition Mobility Steps

Validation Copy memory pages Host to target systems Transfer Turn off Host resources Activate Target resources

The number of Oracle licenses needed does not change before and after the migration

Customer Shared Pool

More SAP performance than any 8-socket system in the industry Comparable to a 128-core, 32-socket Sun M9000

15,600

SAP users on SAP SD 2 Tier

Power 750 Express with DB2

8-core Sun Fire X4270 Xeon 5500

2-sockets

24-core 32-core 32-core HP DL585 Sun T5440 Power 750 AMD

4-sockets

48-core 48-core 128-core HP DL785 Sun x4640 Sun M9000 AMD AMD

8-sockets

BSee SAP Benchmarks chart for detail or SAP website est SAP 2-Tier Results for 2, 4 , 8 and 16 sockets.

32-sockets

The most energy efficient 4-socket system on the planet The first Energy Star certified RISC system

Power 750

Most energy efficient systems Performance Per Watt

Itanium HP rx6600 11 5

SPARC Sun T5440

x86 HP DL585

POWER7 Power 750 with PowerVM

Power 750 Delivers Superior Performance than HP Integrity Entry, Midrange and High End Servers • More than 10X performance than 4-socket HP Integrity rx6600 server • 28% better performance than 64-core HP Integrity Superdome

IBM Power 750 Express

Performance

32 cores

64 cores 32 cores

8 cores Power 750 HP rx6600

16 cores HP rx7640 HP rx8640

HP Superdome

See Power 750 Performance and Efficiency compared to HP Integrity servers for substantiation detail. Source: SPECiint_rate2006. For the latest SPEC benchmark results

POWER7 means more SAP Throughput  SD 2-Tier ERP tests* show on Power 750 show new levels of throughput – Over 7X better per processor throughput than Sun SPARC M9000 – Over 3X better per processor throughput than Sun SPARC T5440 – Over 3X better per processor throughput than HP ProLiant DL585 G6

IBM Power 750 Express

– Over 2X better per processor throughput than Sun X4270

Users per processor/socket

4000 3500 3000 2500 2000 1500 1000 500 0

Pow er 750

SUN M9000

SUN x4270

HP DL585 G6

SUN T5440

Benchmark

3900

3900 Power 750

272 SUN M9000

1180 SUN T5440

1900 SUN x4270

1166 HP DL585 G6

S1

Consolidate HP Integrity rx7640 and 71 Proliant DL380 G5 Linux Servers onto One Power 750 server    

Utilize 3% of the previous energy 97% less space – 3+ racks to one Power 750 552 fewer cores reducing per core s/w licensing Dramatically reduce network and power cabling complexity

HP Integrity rx7640 • 1.6GHz; 10U • 2,128W • SPECint_rate2006: 201

IBM Power 750 Express 71 X HP Proliant DL380 G5 • 3.0Ghz; 2U X 62 servers • 1,193W; Total = 110,949W • SPECint_rate2006: 36.2

 3.55 GHz; 4U  1,950W  SPECint_rate2006: 1060

IBM Power 780 Delivers Performance with Efficiency

780 delivers over 3X the performance per core of HP Superdome and Sun M9000 Performance Per Core Performance Per Watt 780 delivers over 5.8X the performance per watt of HP Superdome and Sun M9000

HP Superdome

Sun M9000

POWER6

POWER7

HP Superdome

Sun M9000

POWER6

POWER7

System

Chip/Core/Thread

Date

SPECint_rate2006

Per core

Maximum energy requirement (WATTs)

Per KWatt

IBM Power 780 (3.8 GHz POWER7)

8/64/256

February 2010

2530

39.5

6,400

395

IBM Power 570 (4.2 GHz POWER6)

16/32/64

October 2008

832

26

5,600

148

Sun SPARC Enterprise M9000

64/256/512

October 2009

2586

10.1

44,800

58

HP Integrity Superdome (1.6 GHz Itanium 2)

64/128/128

September 2006

1648

12.875

24,392

68

Source: http://www.spec.org IBM results available at announcement. All other results as of 01/27/10. Not all results listed. Performance per KWatt is ca lculated by dividing the performance by the recommended maximum power usage for site planning. This defines the requirement for the power infrastructure. Actual power used by the systems will be less t han this value for all of the systems. For HP systems, this information is contained in the QuickSpecs available through www.hp.com. For Sun systems, this information is available through the respective Site Planning Guides available through www.sun.com.

119

Power 750 Express delivers Best of Breed eBS R12 Order To Cash performance Delivers the best eBS R12 Order to Cash Medium 2-tier result. Surpasses 55XX per system, per processor and per core results 64% higher per core and 23% overall against Nehalem 55XX 8-core system.

Template Documentation

IBM Power 750 Express

Power 750 Express delivers Best of Breed eBS R12 Payroll performance Delivers the best eBS R12 Payroll Medium 2-tier result - Surpasses HP per system, per processor and per core IBM Power 750 Express

6 core Power 750 out performs HP’s 8 core DL380 G6 - 49% higher per core and 12% overall Oracle eBS R12 Payroll Batch 257,143 250,000

checks per hour

200,000

257,143

229,885

IBM 750 6-core HP DL380-G6 8-core

198,020

HP DL380-G5 8-core

150,000 114,943 100,000

78,534 42,857

50,000

28,736

0 System Performance

per processor

Template Documentation

per core

19,634

POWER Runs Virtualized Every Time, All the Time • All Power systems run virtualized all the time • VMWare overhead is up to 20% running OLTP applications

VMWare overhead 5 4 3 2 1 0 2 virtual cpu's

4 virtual cpu's ESX

8 virtual cpu's

Native

Source: Virtualizing Performance-Critical Database Applications in VMware® vSphere™ a vailable at http://www.vmware.com/pdf/Perf_ESX40_Oracle-TPC-Ceval.pdf as of August 21, 2009 Relative virtualized performance per core

Database

SAP 2-tier SD

Power 780 TurboCore

Integer

Power 750

Xeon 5500

Fltg Pt Xeon 7400

Run in a virtualized environment & the POWER advantage grows!

OverPOWERing Nehalem More scalability Up to 64 lightning fast cores Avoid scaleout overhead More performance per socket Up to eight lightning fast cores More performance per core Lightning fast cores More systems infrastructure – especially in TurboCore mode Up to 3.5 X the memory per core Up to 7 X the memory bandwidth per core Up to 3.5 X the L2 + L3 cache per core Less virtualization overhead

TPC-C POWER7 vs. Competition (per core results) 160000 140000 120000

POWER7 Nehalem-EX Nehalem-EP Itanium/2 Opteron SPARC(Niagara)

100000 80000 60000 40000 20000 0 TPC-C/Core

www.tpc.org

Active Memory Expansion

Act Mem Exp – Turning a Partition On or Off With HMC, check Active Memory Expansion box and enter true and max memory memory expansion factor To turn off expansion, unclick box Partition IPL required to turn on or off Active Memory Expansion Modeled Statistics: ----------------------Modeled Expanded Memory Size : 8.00 GB Expansion Factor --------1.21 1.31 1.41 1.51 1.61

True Memory Modeled Size -------------6.75 GB 6.25 GB 5.75 GB 5.50 GB 5.00 GB

Modeled Memory Gain ----------------1.25 GB [ 19%] 1.75 GB [ 28%] 2.25 GB [ 39%] 2.50 GB [ 45%] 3.00 GB [ 60%]

CPU Usage Estimate ----------0.00 0.20 0.35 0.58 1.46

Active Memory Expansion Recommendation: --------------------The recommended AME configuration for this workload is to configure the LPAR with a memory size of 5.50 GB and to configure a memory expansion factor of 1.51. This will result in a memory expansion of 45% from the LPAR's current memory size. With this configuration, the estimated CPU usage due to Active Memory Expansion is approximately 0.58 physical processors, and the estimated overall peak CPU resource required for the LPAR is 3.72 physical processors.

127

5.5 true 8.0 max

Sample SAP ERP Workload Test Results Details at Constant Throughput: Single Partition (DB + AppServer) True GB

0%

12%

46%

73%

14.25 12.70 11.25 9.75

Gained GB 0 Total GB

27%

1.55

3.0

4.50

10 5 , 0 0

111% 171%

8.25

6.75

5.25

6.0

7.50

9.0

14.25 14.25 14.25 14.25 14.25 14.25 14.25

Test configs held total memory constant at 14.25 GB, varying mix of real and gained memory. Number cores constant at 4 cores.

Throughput (% Nominal)

% Expand

Throughput

10 0 , 0 0 95,00 90,00 85,00

No Impact on throughput at 111% memory expansion

80,00 75,00

0%

12 %

27%

46%

73%

111%

17 1%

Mem ory Expansion (%)

CPU Utilization Response Time 111% more memory for 15% additional CPU

100

CPU Utilization (%)

93 80

75% more memory for 1% additional CPU

60 60

61

61

61

75

61

0 ,8

0 ,6

Minimal impact to response time at 111% expansion

0 ,4

0 ,2

0 ,0

40

0%

12 %

27%

46%

73%

111%

17 1%

Mem ory Expansion (%)

20

12 8

Response Time (s)

1,0

Your results will vary depending on compressibility of the data and available CPU resource

0 0%

12%

27%

46%

73%

Memory Expansion (%)

111%

171%

Sample SAP ERP Workload, Single Partition With Active Memory Expansion Without Active Memory Expansion + AppServer) (DataBase Partition utilization Partition utilization

 Memory: 100% (18 GB true)  CPU: 88% (12 cores in LPAR) Note: Most of

 Memory: 100% (18 GB)  CPU: 46% (12 cores in LPAR)

the CPU increase is due to additional work done on server

Memory capacity is the bottle-neck  CPU is under-utilized  Handles 1000 simulated users

Higher throughput enabled with the same amount of physical memory  Gain 37% memory capacity  Handles 1700 simulated users

Max Partition throughput: 99 tps

12-core POWER7 partition 18 GB Memory 18 GB true . 0 GB expanded

+ 65%

Max Partition Throughput: 166 tps 12-core POWER7 partition 24.7 GB Memory 18 GB true . 6.7 GB expanded

Expanded Memory Note: This is an illustrative scenario based on using a sample workload. This data represents measured results in a controlled lab environment. Your results may vary.

Active Memory Expansion – Client Deployment Steps Planning Tool

60-Day Trial

Deploy into Production

A. Part of AIX 6.1 TL4 B. Calculates data compressibility & estimates CPU overhead due to Active Memory Expansion C. Provides initial recommendations

A. One-time, temporarily enablement B. Config LPAR based on planning tool C. Use AIX tools to monitor Act Mem Exp environment D. Tune based on actual results

A. Permanently enable Active Memory Expansion B. Deploy workload into production C. Continue to monitor workload using AIX performance tools

Memory Expansion

130

Actual Results App. Performance

Estimated Results

Memory Expansion

Performance

3

CPU Utilization

2

CPU Utilization

1

Time

Memory Expansion

POWER7 Related Announcements

POWER7 Virtualization Support •Maintain 1 to 10 ratio for Physical cores to LPARs  Power 750  Power 755  Power 770 / 78:

Up to 160 (320) LPARS Not Supported Up to 160 (640) LPARs

•Active Memory Expansion  Active Memory Expansion compresses in-memory data to fit more data into memory • Increases the effective amount of memory capacity for AIX partitions

 Managed by the OS and hypervisor • OS compresses and decompress data based on memory accesses

 Is transparent to applications

134

AIX 6 Editions • AIX 6 is available in three different editions: – AIX 6 Standard Edition • Suitable for most UNIX workloads • Vertical scalability up to 64 cores AIX Enterprise Edition

– AIX 6 Enterprise Edition • AIX plus enterprise management • Includes AIX 6 Standard Edition plus Systems Director Enterprise Edition and the Workload Partitions Manager for AIX • Vertical scalability up to 64 cores

AIX Standard Edition

AIX Express Edition

– AIX 6 Express Edition • Lower priced edition targeted a low end servers and consolidation of smaller workloads on larger servers • Includes all the functionality of AIX 6 Standard Edition • Vertical scalability is limited to 4 cores and 8GB of memory per core in a single partition • Clients can use multiple AIX Express Edition partitions in a single larger server

•

Clients can mix multiple AIX editions in the same server

AIX Enterprise Edition AIX Enterprise Edition is a single offering that brings together AIX 6 with key service management capabilities that are designed to:  Improve availability through access to relevant real-time information and predictive monitoring to avoid future problems  Enhance operational efficiency through visualization of resources and centralized deployment and management of virtualized AIX environments  Provide accurate assessment of system resource usage AIX Enterprise Edition includes: – AIX 6 – WPAR Manager V2.1 – Systems Director Enterprise Edition • •

IBM System Director 6.1.2 Active Energy Manager 4.2

• • • • • •

VMControl 2.2 (including Image Management & System Pools) Network Control 1.1 Transition Manager for HP® SIM Service and Support Manager 6.1.2 IBM Tivoli Monitoring 6.2.2 Agents: ITM for Energy Mgmt v6.2; VMControl Agent v2.2; Power System Agents (CEC, AIX Premium, VIOS, HMC) v6.2.1; Systems Director Agent TADDM v7.2

•

AIX Enterprise Edition Helps you Monitor and Repair Power Server Systems  Monitor resources to maintain system availability Platform and enterprise system monitoring Proactive alerts System pool monitoring

 Analyze system status to find the root cause of problems more quickly At-a-glance status of critical systems Root cause analysis Enhanced correlation

 Repair or prevent system faults to reduce service downtime Automated response and updates Workload migration Proactive notification

VMControl Editions: Added Value for PowerVM Clients VMControl Express Edition

VMControl Standard Edition

VMControl Enterprise Edition

Manage resources

Automate virtual images

Optimize system pools

Create/manage virtual machines (x86, PowerVM and z/VM)







Virtual machine relocation







Capture/import, create/remove standardized virtual images





Deploy standard virtual images





Maintain virtual images in a centralized library





VMControl Virtualization Capabilities PowerVM

Create/remove system pools and manage resources in system pools



Add/remove physical servers within system pools



More Easily Migrate Linux x86 Applications to Power Run x86 Linux applications on Power alongside your AIX, i and Linux on Power applications Simplifies migration of Linux x86 applications enabling customers to realize the energy and administration savings of consolidation    

Run most existing 32-bit x86 Linux applications with no application changes Included with the purchase of PowerVM Editions POWER6 blades through Power 595; POWER7 servers – outlook 2Q2010 developerWorks download: http://www.ibm.com/developerworks/linux/pave/ PowerVM™ Lx86 PowerVM x86 x86 Linux x86 Linux App Linux App App Linux Linux Linux

x86 Platforms x86 Platforms x86 Platforms

Install and Run 

No Porting



No Recompile



No changes

x86 Linux App PowerVM Lx86

POWER Linux Application Linux

AIX Application

i Application

AIX

IBM i

Power Systems Platform

140

2009 – 2011 AIX TL Roadmap 10/2009

04/2010

AIX 5.3

10/2010

04/2011

10/2011

POWER7 Hardware Support

TL8 TL9 TL10

TL11 TL11

SP

SP SP TL12

AIX 6.1 TL1 TL2 TL3 TL4 TL4

SP

SP SP TL5 TL6 TL7

TL8 TL0

SP

Service Pack POWER7 Support

AIX 7.1

TL1

TL2

Source for full survey on : http://iticcorp.com/blog/2009/07/itic-2009global-server-hardware-server-osreliability-survey-results/

145

IBM i Roadmap 2008

2009

IBM i 6.1

2010

2011

2012

IBM i Strategy of a major new release of IBM i every two years Next

IBM i

IBM i 6.1.1 IBMSOD i 7.1

„06

V5R2

„07

„08

„09

„10

„11

„12

„13

„14

„15

„16

IBM i Upgrade paths

V5R3

5.4* 6.1* 7.1*

Service

*The projected date for the service of IBM i releases is based on current IBM planning assumptions. Note that it is IBM‟s current practice to support an IBM i release until the next two releases have been made available, plus twenty four months. This slide contains information about IBM‟s plans and directions. Such plans are subject to change without notice.

Partition Mobility

POWER6 POWER6+

POWER7

Binary Compatibility between POWER6 and POWER7 Leverage POWER6 / POWER6+ Compatibility Mode Migrate partitions between POWER6 and POWER7 Servers  Forward and Backward

New DAT320 Tape Drive Up to 100% more

Up to 70% faster Up to 50% less

DAT160 baseline

DAT320 Capacity

DAT320 Speed

DAT320 Energy

Only 30% more

DAT320 Price

• Placed in POWER7 750 (or POWER6 520/550) CEC HH bay

–Available 19 Feb for Power 750, and 16 March for Power 520/550 • Reads/writes DAT160, but does not read/write DAT72 • Requires AIX 5.3, IBM i 6.1, SUSE 10, Red Hat 4.8

(or later)

POWER7 Upgrades

POWER7 System Upgrades – You Can Get there. Upgrades from POWER6 and POWER6+ All existing Power 570 systems can upgrade to POWER7

POWER6+ 570/32 4.2 GHz

Power 780 3.8 GHz / 4.1 GHz

POWER6+ 570

9179-MHB

4.4, 5.0 GHz

POWER6 570

Power 770 3.5 GHz

3.5, 4,2, 4.7 GHz 9117-MMA

9117-MMB

 POWER6 upgrades to POWER7  POWER6+ upgrades to POWER7  Power 570/32 upgrades to POWER7

I/O Upgrade Considerations • All the newer IBM I/O drawers (12X), disk, SSD and PCI adapters used on POWER6 supported on POWER7 servers – May need to move 3.5-inch SAS drives and PCI-X adapters • Older I/O on POWER6 servers, but not on POWER7 servers – RIO/HSL I/O drawers – SCSI disk smaller than 69GB or SCSI drives slower than 15k rpm – QIC tape drives – IOPs and IOP-based PCI adapters (IBM i) 2749, 5702, 5712, 2757, 5581, 5591, 2790, 5580, 5590, 5704, 5761, 2787, 5760, 4801, 4805, 3709, 4746, 4812, 4813 – Older LAN adapters: #5707, 1984, 5718, 1981, 5719, 1982 – Older SCSI adapters: #5776, 5583, 5777 – Telephony adapter: #6412

–See planning web page www.ibm.com/systems/power/hardware/sod2.html

Power SODs for Upgrades Definition “upgrade” as a model change keeping same serial number

• Power 595 – SOD issued in 2009 & augmented 2010

• Power 570 – SOD issued in 2009 – Upgrades announced Feb 2010, shipping June 2010 – Built on unified structure, 9406-MMA must first convert to 9117-MMA • Power 575 and 560 and 550 – SODs not issued

• Power 520 – SOD issued February 2010 with plans to be delivered in 2010 – For Power 520 (8203-E4A) 2-core or 4-core servers – Insight: POWER5 520 to POWER6 520 upgrades did not have savings in the hardware. Client savings were in easy license transfer (including IBM i), documented upgrade procedures for upgrading, and perhaps easier leasing/depreciation structure continuation

POWER7 High-End Statement of Direction • IBM plans to deliver a new high-end server in 2010 with up to 256 POWER7 processor cores • Designed to operate within the same physical footprint and energy envelope of the current 64core Power 595 server. • High-Voltage DC Power option • IBM also plans to provide an upgrade path from the current IBM Power 595 server with 12X I/O to the new POWER7 high-end server.

• Enterprises with multiple systems leveraging PowerVM Live Partition Mobility may use this function to maintain application availability during the upgrade process.

Power Solid State Drives (SSD) Update • Feb 2010 –SSD support in Power 750, 755, 770, 780 system units • Dec 2009 HUGE price action –48% price reduction (plus 30% maintenance reduction) –Aligned SSD with memory price reduction –Some countries deferred price action to 2010 • Oct/Nov 2009 –“Enterprise class vs. consumer SSD” white paper –IBM i analysis tool –SSD configuration & performance enhancements –AIX analysis paper by Dan Braden –#1 SCP-1 benchmark - 595 + 84 SSD

48%