IBM POWER7 pod drobnohledem

Jan Suchy [email protected] +420 737 264 298

IBM POWER7 pod drobnohledem

© 2010 IBM Corporation

IBM Power Systems

Agenda…. 1) IBM POWER7 pod drobnohledem

IBM PowerSystems POWER7 procesor Servery s P7 Výkonnost architektury IBM PowerVM Virtualizace - technologie Efektivní dynamická infrastruktura

2) Virtualizace IBM PowerVM přínosy a úspory v praxi

Konsolidace a centralizace IT Teorie Metodologie Praktická ukázka


IBM PowerSystems

- POWER7 procesor - Servery s POWER7 - Výkonnost architektury


IBM Power Systems

Processor Technology Roadmap

POWER8 POWER7 45 nm

POWER6 65 nm

POWER5 130 nm

POWER4 180 nm

Dual Core Chip Multi Processing Distributed Switch Shared L2 Dynamic LPARs (32)

2001

Dual Core Enhanced Scaling 2way SMT Distributed Switch + Core Parallelism + FP Performance + Memory bandwidth + Virtualization (CPU)

2004

Dual Core High Frequencies Virtualization (Mobility) Memory (SharedPool) Altivec Instruction Retry Dyn Energy Mgmt SMT + Protection Keys DECIMAL FP

2007

Multi Core (8c) On-Chip eDRAM L3 Power Optimized Cores Mem Subsystem ++ (6x) 4way SMT++ Reliability + VSM & VSX (AltiVec) Protection Keys+ DECIMAL FP

Concept Phase

2010 © 2010 IBM Corporation

IBM Power Systems

Výkonnost architektury


IBM Power Systems

POWER, Performance, Virtualization

rPerf per KWatt

30X increase in performance per watt 26X increase in rPerf performance 9+ years of changing the UNIX landscape

DLPAR 1,0 CPU No Virtual I/O

POWER4™ p670 1.1 GHz rPerf: 24.46 KWatts: 6.71

POWER4+™ p670 1.5 GHz rPerf: 46.79 KWatts: 6.71

uDLPAR 0,1 CPU Virtual I/O VIOS Shared pool 2way SMT

POWER5™ p5-570 1.65 GHz rPerf: 68.4 KWatts: 5.2

POWER5+™ p570 1.9 GHz rPerf: 85.20 KWatts: 5.2

uDLPAR 0,1 CPU Virtual I/O + 8Gb, HEA Shared pool, RAM, Ex 4way SMT

uDLPAR 0,1 CPU Virtual I/O + 8Gb, HEA Shared pool, RAM 2way SMT

POWER6™ Power 570 4.7 GHz rPerf: 134.35 KWatts: 5.6

POWER6™ Power 570 4.2 GHz rPerf: 193.25 KWatts: 5.6

POWER7™ Power 780 3.8 GHz


IBM Power Systems

The highest performing 4-socket system on the planet

POWER7 continues to break the rules with more performance SPECint_rate

Itanium HP rx6600

7

SPARC Sun T5440

x86 HP DL585

Power 750

POWER7 Power 750 with PowerVM © 2010 IBM Corporation

IBM Power Systems

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 8

24-core 32-core 32-core 48-core 48-core 128-core HP DL585 Sun T5440 Power 750 HP DL785 Sun x4640 Sun M9000 AMD AMD AMD

4-sockets

8-sockets

•Best SAP 2-Tier Results for 2, 4 , 8 and 16 sockets. •See SAP Benchmarks chart for detail or SAP website http://www.sap.com/solutions/benchmark/sd2tier.epx

32-sockets © 2010 IBM Corporation

IBM PowerVM

- Virtualizace – technologie - Virtual I/O server, NPIV - Efektivní dynamická infrastruktura


IBM Power Systems

Benefits of Using PowerVM on POWER7

Support for 1000 LPARs (VMs) per server – Run up to 160 LPARs on a Model 750 (Feb 2010) – Run up to 320 LPARs on a Model 750 (Oct 2010 SoD)

Increased CPU cores and memory capacity support more powerful virtualized workloads and higher consolidation ratios Superior TCO and ROI compared to competing UNIX and x86-based platforms (including virtualization) Use live partition mobility to migrate workloads from POWER6 to POWER7 platforms with zero downtime


IBM Power Systems

PowerVM Technologies The leading virtualization platform for UNIX, i and Linux enables a more agile and responsive infrastructure

Hypervisor Support for multiple operating environments

Dynamic Logical Partitioning Micro-partitioning, resource movement

Multiple Shared Processor Pools

VIOS

Cap processor resources for a group of partitions

Virtual I/O Server Virtualizes resources for client partitions

Integrated Virtualization Manager Simplifies partition management for entry systems

Power Hypervisor

Shared Memory Pool Memory Expansion

Lx86 Supports x86 Linux applications

Live Partition Mobility Move running AIX and Linux partitions

System Planning Tool Simplifies the planning for and installation of Power servers with PowerVM © 2010 IBM Corporation

IBM Power Systems

PowerVM Virtualization Option Dynamically Resizable

Virtual I/O Server Partition

3 Cores

Int Virt Manager

Linux

6 2 Cores Cores

6 Cores

Storage Sharing Ethernet Sharing Virtual I/O paths

POWER Hypervisor

Web Browser

IBM i AIX V6.1 AIX V5.3

AIX V5.3

AIX V6.1

AIX V6.1

AIX V5.3 Linux Linux

Micro-Partitioning

Offerings for POWER6/7 Standard Enterprise (Partition Mobility) Micro-Partitioning™ Share processors across multiple partitions Minimum Partition: 1/10 processor AIX V5.3 / V6.1 IBM i Linux Virtual I/O Server Shared Ethernet Shared SCSI & Fiber Channel Int Virtualization Manager

IVM

HMC © 2010 IBM Corporation

IBM Power Systems

Shared Dedicated Processors Linux

AIX 5.3 AIX 6.1

AIX 5.3

Linux

AIX 6.1

POWER Hypervisor P P P P P P P P P P P P P P P P

Shared Dedicated / Dedicated Processors

Shared (Non-Dedicated) Processors

Unused capacity in dedicated processor partitions can be “Donated” to

Excess Dedicated Capacity Utilization

shared processor pool Excess cycles will only be utilized by uncapped partitions that have consumed all of their entitled capacity. POWER6 Servers © 2010 IBM Corporation

IBM Power Systems

Micro-Partitioning Technology

Micro-Partitioning technology allows each processor to be subdivided into as many as 10 “virtual servers”, helping to consolidate UNIX® and Linux applications.

Micro-partitions

Dynamic LPARs

Pool of Six CPUs

AIX V6.1

Linux

IBM i

Linux

Partitioning options AIX V5.3

AIX V5.3

AIX V5.2

Whole Processors

– Micro-partitions: Up to 1024*

Configured via the HMC or IVM Number of logical processors – Minimum / Maximum

Entitled capacity Entitled capacity

– In units of 1/100 of a CPU – Minimum 1/10 of a CPU Min Max

Hypervisor

Variable weight – % share (priority) of surplus capacity

Capped or uncapped partitions

Note: Micro-partitions are available via optional Power VM or POWER Hypervisor and VIOS features. © 2010 IBM Corporation

IBM Power Systems

Multiple Shared Processor Pools

P1 A I X

P2 A I X

P3 A I 2X 0.75

Dedicated 2 1 Core Core

P4 L i n 1 u 0.25 x

V Pool: 0 Max Cap: 2 Ent Cap:1

Capped Partition #

Number of VP’s

#

Entitled Capacity

POWER6/7 16-core System

P5 A I X 4

P6 A I X 4

P7 A I X1

P8 A I X2

1.5

0.5

0.25

0.5

V Pool: 1 Max Cap: 10 Ent Cap: 3.25

P9 L i n3 u 0.5 x

P10 I B M2 i 0.25

P11 I B M2

P12 L i n2 u i 0.25 x0.5

P13 P14 P15 7 A A L I I i X3 X1 n 1 u 0.5 0.25 x0.25

V Pool: 2 V Pool: 3 4 Max Cap: 3 Max Cap: 4 Ent Cap: 0.5 Res Cap:0.5 Ent Cap:2

13 Cores ( Shared Processor Pool )

Capping at the pool level Over commit processor resources © 2010 IBM Corporation

IBM Power Systems

Active Memory Expansion & Active Memory Sharing Active Memory Expansion

Active Memory Sharing

Effectively gives more memory capacity to the partition using compression / decompression of the contents in true memory AIX partitions only

Moves memory from one partition to another Best fit when one partition is not busy when another partition is busy AIX, IBM i, and Linux partitions 15

10

5

0

Active Memory Expansion

#10 #9 #8 #7 #6 #5 #4 #3 #2 #1

Active Memory Sharing

Supported, potentially a very nice option Considerations – Only AIX partitions using Active Memory Expansion – Active Memory Expansion value is dependent upon compressibility of data and available CPU resource


IBM Power Systems

Sample SAP ERP Workload, Single Partition (DB + AppServer) Without Active Memory Expansion

With Active Memory Expansion

Partition utilization

Partition utilization – Memory: 100% (18 GB true) – CPU: 88% (12 cores in LPAR)

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

Note: Most of the CPU increase is due to additional work done on server

Memory capacity is the bottle-neck

Higher throughput enabled with the same amount of physical memory

– CPU is under-utilized – Handles 1000 simulated users

Max. Partition throughput: 99 tps

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

– Gain 37% memory capacity – Handles 1700 simulated users

+ 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. © 2010 IBM Corporation

IBM Power Systems

Sample SAP ERP Workload, Multiple Application Partitions Without Active Memory Expansion


Partition utilization

Partition utilization – Memory: 100% (48 GB true) – CPU: 94% (24 cores in 3 8 core LPARs)

– Memory: 100% (48 GB) – CPU: 76% (24 cores in 3 8 core LPARs)

Note: Majority of the CPU increase is due to additional work done on server

Memory capacity is the bottle-neck – CPU is under-utilized – Handles 2900 simulated users – Partitions have reached physical memory limitations by showing moderate paging

Max. Partition throughput: 286 tps

3 x 8-core POWER7 partitions 48 GB Memory 48 GB true . 0 GB expanded

+ 35%

Higher throughput enabled – Gain xx% memory capacity – Handles 4090 simulated users – Partitions showing no paging

Max. Partition Throughput: 385 tps

3 x 8-core POWER7 partitions xx GB Memory 48 GB true . xx 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. © 2010 IBM Corporation

IBM Power Systems

Sample SAP ERP Workload, Enabled Additional Application Partition Without Active Memory Expansion


System Utilization

System Utilization

– Memory: 100% (48 GB) – CPU: 94% on 32 cores

– Memory: 100% (48 GB) – CPU: 76% on 24 cores 25% (8 core) unused

Note: Majority of the CPU increase is due to additional work.

Higher throughput enabled

Memory Capacity is the bottle-neck

– CPU is under-utilized – Handles 2900 simulated users – Partitions have reached physical memory limitations by showing moderate paging

– Enabled unused CPU resources (25% of server) with no add’l physical memory. – Gain 30% in application server memory capacity – Handles 5000 simulated users

+ 60%

System Throughput: 286 TPS

3 x 8-core POWER7 partitions

System Throughput: 460 TPS

4 x 8-core POWER7 partitions

LPAR 4 (AppServer)

LPAR 3 (AppServer)

LPAR 2 (AppServer)

LPAR 1 (DB + App)

48 GB true . 14 GB expanded LPAR 4 (IDLE)

LPAR 3 (AppServer)

LPAR 2 (AppServer)

LPAR 1 (DB + App)

48 GB true . 0 GB expanded

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. © 2010 IBM Corporation

IBM Power Systems

Active Memory Expansion - CPU & Performance 2 1

% CPU utilization for expansion

1 = Plenty of spare CPU resource available

Very cost effective

2 = Constrained CPU resource – already running at significant utilization

Amount of memory expansion There is a knee-of-cure relationship for CPU resource required for memory expansion – Busy processor cores don’t have resources to spare for expansion – The more memory expansion done, the more CPU resource required

Knee varies depending on how compressible memory contents are © 2010 IBM Corporation

IBM Power Systems

AIX Workload Partitions Separate regions of application space within a single AIX image Improved administrative efficiency by reducing the number of AIX images to maintain Software partitioned system capacity –Each Workload Partition obtains a regulated share of system resources –Each Workload Partition can have unique network, filesystems and security

Two types of Workload Partitions –System Partitions –Application Partitions

Workload Partition Test

Separate administrative control –Each System Workload partition is a separate administrative and security domain

Shared system resources

Workload Partition Billing

Workload Partition Application Server

Workload Partition BI

Workload Partition Web Server

Workload Partition Test

Operating System, I/O, Processor, Memory

AIX 21



IBM Power Systems

AIX Live Application Mobility Move a running Workload Partition from one server to another for outage avoidance and multi-system workload balancing

Workload Partition App Server

Workload Partition Web

Workload Partition Data Mining

Workload Partition e-mail

Workload Partition QA

Workload Partition Dev

Workload Partition Billing

AIX Workload Partitions Manager

Policy

AIX

Works on any hardware supported by AIX 6, including POWER5 and POWER4 22



IBM Power Systems

Virtual I/O Server


IBM Power Systems

PowerVM Virtual I/O Server (VIOS)

Workload management and provisioning

AIX 5 partitions

Linux partitions

AIX 6 AIX 6

AIX 6

Global Instance

e-

Workload Partition mail

Global Instance


e-

Workload Partition mail


Workload Partition

Workload Partition

QA Workload Partition Billing

Hardware Management Console

Hypervisor


i5/OS ®

QA Workload Partition Billing

Virtual I/O servers

Unassigned on demand resources


Virtual processors

Virtual Network

Virtual adapters

Processors Service processor

Memory Expansion slots Local devices & storage Networks and network storage


IBM Power Systems

Virtual I/O Server - Value Proposition “Your new system will be ready in …” “20 Minutes” or “20 Days”

Economical I/O Model 10% 10% 10% 10% 10% 10% 10% 10%

}

40% 40%

Quick Deployment

Virtual

SAN + Net Cables Switches Ports + Power

Reduced Infrastructure

Server Consolidation © 2010 IBM Corporation

IBM Power Systems

Virtual I/O Server Home Website

Download Latest Readme

Master Datasheet Red Books

Online Manual & Command Ref.

White Papers https://www14.software.ibm.com/webapp/set2/sas/f/vios https://www14.software.ibm.com/webapp/set2/sas/f/vios/documentation/home.html


IBM Power Systems

Your Reference Library


IBM Power Systems

Processor Virtualisation Evolution (CPU Sharing)

Old Style Separate Systems

LPAR Server Consolidation LPAR size via start time boundaries

DLPAR Dynamic live boundary changes Manual or scripts

SPLPAR Shared Processor automatic adjusts at millisecond level by Hypervisor

Pre-2000

~2001

~2002

~2005


IBM Power Systems

Processor Virtualisation Evolution (CPU Sharing)

SPLPAR Shared Processor automatic adjusts at millisecond level by Hypervisor

Harvesting “Spare” capacity ready for adding more workloads at zero hardware cost

Partition Mobility Make a cluster of your machines & flow your workload between them

~2005

~2006

~2008


IBM Power Systems

Virtual I/O Server (Adapter Sharing)

The I/O centric view of the world: CPU used to “modify & feed data” to the networks & disks

Network

Disk


IBM Power Systems

VIOS

Pure Virtual VIOS

Disk

VIOS

Network

Network

VIOS

Disk

Virtual except high I/O LPARs

Disk

Pure Direct

Production Direct the rest Virtual

Network

Disk

Network

Virtual I/O Server (Adapter Sharing)

VIOS

Reduced Cost © 2010 IBM Corporation

IBM Power Systems

Where are You? A

B

C

D

E

F

1

2

3

4


VI O S

IBM Power Systems

Storage Virtualisation

With NPIV

VIO client

VIO client

Note

Generic SCSI disk

EMC 5000 LUN

vSCSI

FC Adapters

IBM 2105 LUN Virtual FC Adapters

Virtual SCSI Adapters SCSI SAS

VIOS

VIOS

Storage Virtualiser

2. 1

FC Adapters

Pass Through mode

VIOS Admin in charge SAN

EMC 5000 LUN

IBM 4700 LUN

NPIV SAN

EMC 5000 LUN

SAN Admin Back in charge

IBM 4700 LUN © 2010 IBM Corporation

VI O S

IBM Power Systems

NPIV What you do?

1.

2. 1

HMC 7.3.4 configure – – –

Virtual FC Adapter Just like virtual SCSI On both Client and Server

Virtual I/O Server © 2010 IBM Corporation

IBM Power Systems

NPIV What you do?

2.

VI O S

2. 1

Once Created: LPAR Config Manage Profiles Edit click FC Adapter Properties and the WWPN is available


VI O S

IBM Power Systems

Previous Virtual SCSI to Fibre Channel VIOS

AIX

LVM

LVM multipathing

fibre channel HBAs

LVM multipathing

AIX MPIO

Disk Driver

Disk Driver

Disk Driver

VSCSI target

VSCSI HBA

VSCSI HBA

2. 1

VSCSI target

fibre channel HBAs

PHYP

SAN © 2010 IBM Corporation

IBM Power Systems

New NPIV pure Fibre Channel VIOS

AIX

VI O S

LVM

2. 1

Storage Multipathing Disk Driver

fibre channel HBAs

passthru module

VFC HBA

VFC HBA

passthru module

fibre channel HBAs

PHYP

SAN No VIOS side multipath, more client setup per LPAR but Thinner Stack


VI O S

IBM Power Systems

Heterogeneous Multipathing

VIOS

AIX Client LPAR

Passthru module

Fibre HBA

Fibre HBA

NPIV

NPIV

A

2. 1

Note: Multipath FC & FC

POWER Hypervisor

Storage Controller SAN Switch

Note: NOT vSCSI

SAN Switch

A

B

C

D

A’ ’

B’

C’

D’


Jan Suchy [email protected] +420 737 264 298

Virtualizace IBM PowerVM - přínosy a úspory v praxi


Konsolidace a centralizace IT

- Teorie - Metodologie - Praktická ukázka


IBM Power Systems

Situace: IT systémy a infrastruktura dosahují kritických hodnot

Dramatické bujení serverů a síťových zařízení Nadměrné užívání energií a vytváření tepelných ztát Nepřiměřené napájení a chlazení infrastruktury Datová úložiště a jejich synchronizace Předpoklad toho, že „všechno“ je připojeno Lineární náklady na IT obsluhu Prudce rostoucí náklady za licence SW Nevysvětlitelné výpadky

Mezitím, očekávání zákazníků, tlak konkurence, regulační opatření a tlaky na fiskální výsledky neustále rostou. © 2010 IBM Corporation

IBM Power Systems

Proč virtualizovat?

Efektivní využití zdrojů Zjednodušení správy a údržby Snížení množství HW vybavení Dynamické změny HW parametrů Nové služby bez nákupu infrastruktury

Snížení energetických a prostorových nároků

Snížení nákladů © 2010 IBM Corporation

IBM Power Systems

Typická malé vytížení serverů Typický UNIX nebo x86 server běžící jednu úlohu a operační systém je vytěžován na 10 - 20% 100%

Konfigurace pro plánovaný růst

(20% nepoužito)

80%

Kolik jsme zaplatili…

60%

Konfigurace plánované pro špičky

(50% nepoužito)

Systémy čekající na I/O a paměťový přístup i u prac. systémů (20% unused?)

40% 20% 0%

Kolik dostáváme…

Výsledkem je, že 80% hardware, software, údržbou, místem, a energiemi za které jste zaplatili, je plýtváno. © 2010 IBM Corporation

IBM Power Systems

Náklady na vlastnictví serverů

Spending (US$B)

Installed Base (M Units)

$300

50 45

$250

Náklady na napájení a chlazení x8

$200

Náklady na administraci a management x4 Náklady za pořízení nových serverů

40 35 30

$150

25 20

$100

15 10

$50

5 $0

20 10

20 09

20 08

20 07

20 06

20 05

20 04

20 03

20 02

20 01

20 00

19 99

19 98

19 97

19 96

0

Source: IDC, Virtualization 2.0: The Next Phase in Customer Adoption


IBM Power Systems

Finanční přínos konsolidace / virtualizace

Náklady Údržba

Software

Personál

Možné úspory

Hlavní součásti

69% - 76%

Menší počet nových serverů redukuje náklady na údržbu.

65% - 69%

31% - 45%

FINANCIAL SERVICES COMPANY 59.3

Conventional Virtualized

22.3 (62% less)

Menší počet kopií, méně procesorů vede k nižšímu počtu licencí a tím k nižší podpoře a poplatkům za údržbu software.

MANUFACTURING COMPANY

Redukce počtu serverů a propracovaná technologie managementu virtuálních prostředí vede k redukci požadavků na administrátory systémů.

RETAIL COMPANY

25.3

Conventional Virtualized

9.9 (61% less)

Conventional

21.3 9.7 (55% less)

Virtualized 0

Vybavení

52% - 61%

Menší počty serverů, menší rozměry a váha, energeticky efektivnější technologie zajistí menší požadavky na místo a energie v datových centrech.

Maintenance System software Database software Personnel Facilities

10

20

30

40

50

60

$ Millions Technology Group, February 10, 2007. Study methodology: Companies in financial services, manufacturing and retail with $15 Billion+ revenues and total 200,000+ employees focusing on UNIX large enterprise environments with multiple, broad-ranging applications. Study compared the cost of the company's workload running on multiple vendor servers and employing minimal virtualization to the cost of the company's workload. This cost analysis was performed for financial services, manufacturing and retail example environments with an overall average savings of up to 62% in TCO savings by virtualizing. TCO depends on the specific customer environment, the existing environments and staff, and the consolidation potential. © 2010 IBM Corporation

IBM Power Systems

Historie serverové virtualizace

Již v roce 1961 je předveden Compatible Time Sharing System (CTSS), umožňující sdílení procesorového času (time sharing) – první VM (virtuální stroj). 1964: rok uvedení platformy mainframe na trh, počátek vývoje operačního systému CP-40, který nabízí každému uživateli kompletní server za pomoci virtualizace prostředků. 1969: CP-67 verze 2, 1970: CP-67 verze 3; tento operační systém je provozován ve 44 datových centrech. 70. léta: přichází nová generace mainframů. Do roku 1989 prodáno více než 20 000 licencí VM. 90. léta virtualizace se objevuje v unixovém světě. Po roce 2000 virtualizace vstupuje do x86 světa.


IBM Power Systems

Kdy virtualizovat?

Starý přřístup: Každý server je dimenzován na špičky požadovaného výkonu, mnohé jsou předimenzované 100

100

100

100

100

80

80

80

80

80

60

60

60

60

60

40

40

40

40

40

20

20

20

0

0

0

+

20

+

0

500 Capacity = Units

20 0

Server 1

Server 2

Server 3

Server 4

Server 5

CPU Capacity Units = 100





Návrh serveru odpovídá maximálnímu současnému zatížení.

server hosující několik virtualních serverů 250

C PU C apacity U nits

Nový přřístup: Server s dynamickými virtuálními servery, které umožní automatický přesun procesorů.

Server 5

200

Server 4 150

Server 3

100

Server 2

50

Server 1

0 1

2

3

4

5

6

7

8

9

10

11

12

minute time intervals, stacked CPU utilization


IBM Power Systems

Vliv virtualizace na procesory

Jeden fyzický server: Single Appliation Server (2 CPUs) Průměrné vytížení: 20.7% 80% Špičkové vytížení: 79% 70% 60% Více virtualních serverů zvyšuje průměrné vytížení 50% ale špičkové se příliš nemění: 40% 30% 8:1 průměrné: 39%, špičkové 76% 20% 16:1 průměrné: 48%, špičkové 78% 10% 0% 64:1 průměrné: 61%, špičkové 78% Počet potřebných procesorů roste pomaleji než počet přidávaných serverů.

16 to 1 Consolidation (12 CPUs)



80%

80%

80%

70%

70%

70%

60%

60%

60%

50%

50%

50%

40%

40%

40%

30%

30%

30%

20%

20%

20%

10%

10%

10%

0%

0%

0%


IBM Power Systems

Koncepty virtualizace serverů Hardware Partitioning Apps

...

Apps

Apps

OS

OS

OS

Partition Controller

Bare Metal Hypervisor

Adjustable partitions

...

Hosted Hypervisor

Apps

Apps

OS

OS

...

Apps OS

Hypervisor Hypervisor Host OS

SMP Server Server is subdivided into fractions each of which can run an OS

Board-level partitioning S/370 SI->PP & PP->SI, Sun Domains, HP nPartitions Core/thread-level partitioning Original POWER4™ LPAR HP vPartitions Sun Logical Domains

SMP Server Hypervisor provides fine-grained timesharing of all resources

Hypervisor firmware z®

System PR/SM and z/VM PowerVM hypervisor

Hypervisor Software VMware ESX Server Xen Hypervisor Microsoft Hyper-V

SMP Server Hypervisor uses OS services to do timesharing of all resources

Hypervisor software runs on a host operating system VMware GSX Microsoft Virtual Server HP Integrity VM User Mode Linux® Linux KVM

častá kombinace třetí možnosti s první nebo druhou © 2010 IBM Corporation

IBM Power Systems

PowerVM poráží VMware ve výkonu AIM7 PowerVM vs vSphere 4.0 Single VM Scaling 140000 120000

jobs/min

100000

-52% -42%

80000

PowerVM

-43%

60000

vSphere 4.0 -41%

40000 -36%

20000 0 1vcpu

2vcpu

4vcpu

6vcpu

8vcpu

Number of virtual processors P550 8cores, 5GHz, 64GB RAM, PowerVM HP DL370 Intel Xeon 5550 2.9T GHz (Turbo Mode) 8 cores, 128GB RAM (HT enabled in BIOS, Intel VTx with EPT HW Virtualization Assist) Red Hat 5.3 (x86_64) with AIM7 Compute Benchmark © 2010 IBM Corporation

IBM Power Systems

PowerVM poráží Hyper-V ve výkonu AIM7 PowerVM vs Windows HyperV VM Scaling 140000 120000

jobs/min

100000 -89%

80000 -86%

PowerVM

60000 -75%

HyperV

40000 -77%

20000 0 1VM

4VMs

6VMs

8VMs

Number of Virtual Machines P550 8cores, 5GHz, 64GB RAM, PowerVM HP DL370 Intel Xeon 5550 2.9T GHz (Turbo Mode) 8 cores, 128GB RAM (HT enabled in BIOS, Intel VTx with EPT HW Virtualization Assist) RHEL 5.3 (GA, x86_64) with AIM7 Compute Benchmark © 2010 IBM Corporation

IBM Power Systems

Možnosti virtuálních serverů

Statické virtuální servery – Změna charakteristik vyžaduje restartování OS Dynamické virtuální servery Zátěž virtuálních serverů – CPU, zlomky CPU v průběhu 24 hodin – RAM – Bez nutnosti restartování OS SAP SAP SAP CRM R/3 TEST Virtualizace IO PROD PROD DEV – Disky – LAN Automaticky reagující (dynamické) virt. servery – CPU – RAM Live partition – Uživatelsky definovaná pravidla mobility Mobilita virtuálního serveru – Zachování běhu aplikace


IBM Power Systems

Příklad: virtualizace a mobilita

Požadavek: Více výkonu pro ‚modrou‘ databázi. Zachovat zdroje ‚červené‘ a ‚zelené‘ databáze. Zachovat běh ostatních databází (možné snížit výkon).

SAP APP SAP DB SAP CRM

SAP QA

SAP DEV

SAP BW

Outsourcing pro finanč©n2010 í instituce , EU IBM Corporation

IBM Power Systems

Příklad: virtualizace a mobilita Požadavek: Konsolidace cca 350 serverů na 20 fyzických serverech. Libovolný virtuální server je spustitelný na libovolném fyzickém serveru. Tedy cca 7000 profilů pro virtuální servery a jejich přesuny. Desítky tisíc virtuálních LAN, disků apod.

Státní správa, Skandinávie © 2010 IBM Corporation

IBM Power Systems

Metodologie konsolidace


IBM Power Systems

Metodologie konsolidace – IBM Zodiac

Zachytit stávající stav >>> 4) Náklady / údržžba, prostřředí, SW, podpora

1) Soupis IT systémů ů

2) Aplikace / funkcionalita

3) Hranice datacentra


IBM Power Systems

Metodologie konsolidace – IBM Zodiac výsledek


IT Resource Optimisation Zodiac v6.07 Jan Suchy

Zodiac Infrastructure Study prepared for Financial Services Company


IBM Power Systems

Confidentiality and Liability

The data used to generate this report is primarily based on IBM experience and available industry data. This report is intended to illustrate the kinds of potential benefits that may be achieved by some customers through infrastructure simplification. This does not mean that in your or any other customer consolidation that such benefits will be achieved. Potential benefits depend, among other factors, on the specific customer environment, the existing customer environment and staff, and the consolidation potential. IBM is providing this report to you on an AS IS basis without warranties of any kind. In no event will IBM be liable to you for any direct, indirect, special or other consequential damages for any use of this report. Any configurations contained in this report are provided as samples by IBM representing typical solutions for consolidation. Whilst each configuration has been reviewed by the IBM configuration tool for accuracy in a specific situation, they are intended to illustrate typical capacity & costs, and there is no guarantee that the configurations will integrate into the customer's operational environment. If there are concerns about any elements of a proposal then a Solution Assurance Review should be conducted. The entire IDEAS International dataset is used under their copyright and may not be supplied or included in its entirety in any client report; however, a small sample may be used in order to illustrate the internal sizing process used to select target servers for the client. © 2010 IBM Corporation

IBM Power Systems

Business Case Summary

Client:FSC Total initial investment in in storage & server technology to initiate a simplified architecture: CZK 29 022 636 Our illustration shows a potential reduction of recurring costs. Business Case Period: 4 years. Base Case: CZK 188 579 860 Alternate Case: CZK 105 549 628 Potential Cost Reduction: CZK 83 030 232 Estimated Payback Period: 1yr 4m The total investment cost does not include write-off charges for current assets. Migration Cost: Not included Please refer to backup slides for further details of computations.

Cumulative Cost 250 000 000.00

200 000 000.00

150 000 000.00

100 000 000.00

50 000 000.00

0.00 E.O.year 0

year 1

year 2 C urrent

year 3

year 4

Alt.C ase


IBM Power Systems

p7-770(48)3.1 116 ud16 Ostatní Aplikace

Grand Total

Average of new Utln

Average of old Utln

New RIPS

Old RIPS

Old Load

New #Cores

Old #Cores

New #CPU

Old #CPU

New Images

Old Images

New Servers

Old Servers

FullName

new-mod

Target Servers by Solution Group

13

2 17 14 129 12 174 96 45 347.2 90 694.3 113 643.8 50% 40%

13

2 17 14 129 12 174 96 45 347.2 90 694.3

13

2 17 14 129 12 174 96 45 347.2 90 694.3 113 643.8 50% 40%

113 643.8 50% 40%

Reduction of Physical Servers: 13 : 2 © 2010 IBM Corporation

IBM Power Systems

Increase Energy Efficiency (for Available Capacity)

85% less servers 25% more capacity 78% less electricity 120

114

18.000

16.201

16.000

100 91

14.000

80

12.000 10.000

60 8.000 40

6.000

32

4.000 20

13

2.857 7

2.000

2 -

C urrent #Physical Servers

Alt.C ase Total RIP C apacity

Systems kW

RIPs/Watt

Energy Efficiency Improvement: 467% © 2010 IBM Corporation

IBM Power Systems

116: Ostatní Aplikace Ostatní Aplikace

Capacity Server Type Total Cores Used Cores Total CPUs #Logical Servers #Physical Servers Ave.Log.Srv RIP Total RIP Capacity Total RIP Workload Ave CPU %Utilisation

Target Utilisation 0 Decimal Places

Annual Operating Costs (AOC) Staff Cost Code Software Cost Code Software Cost /CPU Software Cost /Lsrv Software Cost /Psrv Software M&S Hardware Maint Space Power Staff Cost Depreciation Total AOC est.potential saving /yr One Time Costs (OTC) Software Purchase Hardware Purchase Transition Total OTC Write-Off Net Cash Investment Year Projection OTC + 4x AOC 0yr saving Payback Period Project

Current 174.00 174.00 129.00 17.00 13.00 5 335.0 90 694.3 45 347.2 50.00%

Actual AltCase2 7:1 p7-770(64)3.1 128 69 16 14 2.00 8 167.6 114 345.9 45 347.2 39.66%

AltCase1 7:1 p7-770(48)3.1 96 66 12 14 2.00 8 117.4 113 643.8 45 347.2 39.90%

Unix aixF5.oraEE 1 802 880 0 0 15 549 840 4 649 664 18 286 667 279 1 043 280 0 21 928 349 25 216 616

Unix aixF5.oraEE 1 802 880 0 0 14 873 760 3 487 248 13 714 585 558 1 043 280 0 20 003 560 27 141 405

392 000 38 304 848 0 38 696 848 0 38 696 848

294 000 28 728 636 0 29 022 636 0 29 022 636

121 760 580 66 819 280 1yr 6m

105 549 628 83 030 232 1yr 1m

Year Projection Thousands

116: ud16

200 000 180 000 160 000 140 000 120 000 100 000 80 000 60 000

Unix unix/16 182 915 0 0 23 596 000 20 132 000 263 429 1 745 936 1 407 600 0 47 144 965

0

188 579 860 Time 0yr 0m

40 000 20 000 0 1 Transition Software Purchase Staff C ost Power Software M&S

2

3

Hardware Purchase Depreciation Space Hardware Maint


IBM Power Systems

116: Ostatní Aplikace Environmental Analysis

Power & Space

Current 165.0 3.9

Alt.Case.2 32.0 0.8

Alt.Case 24.0 0.6

70.0

Total RackU Racks (42/40U utilised) Systems kW Distribution kW Mechanical kW Total kW

31.7 5.7 26.0 63.5

9.5 1.7 13.0 24.3

7.0 1.3 13.0 21.3

40.0

60.0 50.0

30.0 20.0 10.0

Energy Efficiency Relative RIPs /Watt Watts/Log.Srv Power Cost per Logical Server

0.0

1.0 3484.4 102 772

3.3 8786.3 47 695

4.5 8732.3 41 854

150.3 27.1 123.3 300.6 6.6

45.1 8.1 61.6 114.9 2.5

33.2 6.0 61.6 100.8 2.2

1 Systems kW

2 Distribution kW

3 Mechanical kW

CO2 Emission Systems Distribution Mechanical tonnes CO2 / yr tonnes CO2 / kRIP


Otázky a odpovědi…

Dekuji za pozornost_


IBM POWER7 pod drobnohledem

Recommend Documents