i/o for virtualized multi-core servers

1

I/O for Virtualized Multi-Core Servers

Iddo Kadim, Intel CorporationDirector, Virtualization Technology

Server Platforms GroupIntel Corporation

2* Other names and brands may be claimed as the property of others. Copyright © 2008, Intel Corporation. All rights reserved

Agenda

The value of multi-core computing

The virtualized server

Optimizing I/O in the virtualized server


Genomics Research Medical Imaging

Weather Prediction

Financial Analysis

Oil Exploration

Design Simulation Cloud Computing Data Center Refresh

Digital Content

Social Networks

Insatiable Demand for Servers

1980 1985 1990 1995 2000 2005 2010

UNITS

http://www.apple.com/science/software/lifescience.html

http://www.forecastsforfilms.com/HurricaneFromSpace2-BPSPP-Ed.jpg

4* Other names and brands may be claimed as the property of others. Copyright © 2008, Intel Corporation. All rights reservedSource: IDC.

Power & CoolingFor every $1 spent on server HW, 50 cents spent on power/cooling

Operating Cost$8 in maintenance for every $1 spent on new infrastructure

Space CrunchDatacenter construction costs>$1,000 /sq ft

Server Sprawl>30m physical servers currently installed: growing 4X next 10 years

IT Constraints on Growth

Source: IDC, "Virtualization Across the Enterprise: Progressing From Cost Reduction to Business Growth," doc # VFE2008_04, May 2008.

* Other names and brands may be claimed as the property of others. Copyright © 2008, Intel Corporation. All rights reserved


Intel Multi Core Intel Single Core

0

2

4

6

8

10

12

14

2002 2003 2004 2005 2006 2007 2008

Impact of Multi-Core R

ela

tive P

erf

orm

an

ce

Quad Core65nm 45nm

Dual Core

Quad-Core Benefits a Broad Range of Workloads

For notes and disclaimers, see legal information slide at end of this presentation.


2004• 5.1M bops• 6 racks• 126 servers• 240 sq ft• 48 kW

Server Compression

EstimatedEnergy Savings

$53KSavings3

AnnualEnergy Costs

87%Reduction2

Floor Space

83%Reduction1

The Bottom Line Benefits for IT (<2 year ROI4)

2008• 5.1M bops• 1 rack• 17 servers• 40 sq ft• 6 kW

For notes and disclaimers, see legal information slide at end of this presentation.

Source: Intel January 16, 2008. Performance comparison using SPECjbb2005 bops (business operations per second) between (2004) 2 socket single core Intel Xeon processor (3.6GHz) and (2008) 2 socket Quad-Core Intel Xeon processor E5450 (3.0GHz, 80W, 1333MHz) measured 8/22/07. 1 Floor space based foot on sq ft reduction. 2 Energy costs based on total solution rack power. 3Energy Savings based on an electric rate of $0.10/kWh assuming 33% average server load. 4 ROI is calculated based on cost of new servers ($6,264 per server pricing based on HP DL 380G5 (32GB RAM) as of Nov 17, 2007. source www.hp.com) divided by energy savings per year. Actual performance results and savings may vary depending on configuration. See backup for details.


Evolution of Virtualization

Virtualization 1.0

Managing IT and Capital Expenses

Test & Development

Test

HWVMM

OS

App

OS

App

Production

HWVMM

OS

App

HWVMM

OS

App

HWVMM

OS

App

HWVMM

OS

App

OS

App

OS

App

ServerConsolidation

Virtualization 2.0

Flexible Workload ManagementHWVMM

OS

App

OS

App

HWVMM

OS

App

OS

App

High Availability/ Disaster Recovery

Load Balancing

HWVMM

OS

App

HWVMM

OS

App

OS

App

HWVMM

OS

App

OS

App

HWVMM

OS

App

Workload 1 Workload 2

Workload 1 Workload 2

“Virtualization 3.0”

Adaptive Continuity


1.00

1.13 1.17 1.18 1.20 1.21

1.38

SPECWeb*2005

SPECfp*_rate2006

SPECint*_rate2006

Vmmark* -

VMware* ESX

SPECjbb*2005

SPECjbb*2005 -

PPWImportance of Native Performance

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. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, visit http://www.intel.com/performance/resources/limits.htm property of others.

Data Source: Published, measured, submitted or approved results as of April 7, 2008. See backup for details;

Native Performance Drives Virtualized Performance

Baseline Intel®

Xeon® 5300 Series

Intel® Xeon® 5400 series:Up to 21% gain in performance

& Up to 38% gain in Energy Efficiencyvirtualization performance is

proportional to native performance

‘Native’

‘Virtual’

Overhead %


Virtualization Hardware Assists

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. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, visit http://www.intel.com/performance/resources/limits.htm

Intel® Xeon® X7350

1.00

1.35

Intel® VTFlexPriority Off

Intel® VT FlexPriority On

Example:Intel® VT FlexPriority – Virtual

Iron 4.0.2 on vConsolidate1

Hardware Assists improve virtualization performance & efficiency

1. 32-bit OS include Microsoft Windows* XP, Windows 2000 Server up to Windows Server 2003 SP1

2. All measurements conducted internally on platforms running 4 x Quad-Core Intel® Xeon® Processor X7350, 32GB memory, vConsolidate Beta 2, Virtual Iron 4.0.2 software. 1 CSU configuration, Sept 2007. Boot time improvement measured on Windows XP. vConsolidate measurement conducted by configuring the system with Windows 2000 SP4

‘Native’

‘Virtual’

Overhead %

virtualization performance is proportional to native performance


Intel® VT for Directed I/O (VT-d)

• VT-d provides support for DMA and interrupt remapping• Supports I/O device assignment to VMs, security, etc.

CPU CPU

DRAM

South Bridge

System Bus

PCI ExpressPCI, LPC, Legacy devices, …

IntegratedDevices

North Bridge

VT-dPCIe* Root Ports


VMMVirtual Device Emulation

Intel® VT for Directed I/O (Intel® VT-d)VM

BufferDriver A

Application

VM

VMM

I/O Device HardwareDevice A

|||||||||||||||||||

Chipset

Device A|||||||||||||||||||

BufferDriver A

Emulation Based Virtual I/O Direct Assigned I/O with Intel® VT-d

BufferDriver A

ApplicationOS OS

Intel VT-d Remap

Native Guest OS driver

No HW changes

Eliminates intermediate paths and VMM overheads

Phys MemIO Devices

VT-

d

Chipset

Better reliability through device isolation

Protected Memory regions


Evolution of I/O Virtualization

NIC

VM1

vswitch

.…

VMM

VM2 VMn

The Future: VM mobility with direct assigned virtual resources

Hypervisor SW – EmulationHigh performance tollCommon denominator I/O

HW Assists – queuingNear native I/O performanceCommon denominator I/O

Direct assignment of shared virtual resourcesDedicated, Isolated I/O Native performanceScalability


Storage, Network and Virtualization

Fibre Channel

Fibre Channel

iSCSI

iSCSI

0%

10%

20%

30%

40%

50%

60%

General Servers

Virtualized Servers

18% (2)

SAN Attach Rates

54% (1)

VMM

EDGE SWITCH

CORESWITCH

FC

VM1 VM2 VM3 VMn

VIRTUALSWITCH

FC Director

FC Edge

1GbE

LAN SAN

1 Source: IDC Multiclient Study, “Server Virtualization on the Move 2007: Foundation for the Dynamic Data Center,” July 20072 Source: Intel estimates

Simplified, Dynamic, and Power Efficient


Legal Disclaimer• INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL® PRODUCTS. NO

LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL® PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. INTEL PRODUCTS ARE NOT INTENDED FOR USE IN MEDICAL, LIFE SAVING, OR LIFE SUSTAINING APPLICATIONS.

• Intel may make changes to specifications and product descriptions at any time, without notice.• All products, dates, and figures specified are preliminary based on current expectations, and are subject to

change without notice.• Intel, processors, chipsets, and desktop boards may contain design defects or errors known as errata, which

may cause the product to deviate from published specifications. Current characterized errata are available on request.

• 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.

• Intel, Intel Inside, Core, Xeon, Intel Virtualization Technology, FlexPriortiy and the Intel logo are trademarks of Intel Corporation in the United States and other countries.

• *Other names and brands may be claimed as the property of others.• Copyright © 2008 Intel Corporation. All rights reserved.


Backup


“Impact of Multi-Core ”

Intel Pentium® III Microarchitecture Intel NetBurst® Microarchitecture Intel Core® Microarchitecture

Legal information slide

Results are Published/Measured/Submitted/Approved as of Nov 12, 2007.

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. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, visit http://www.intel.com/performance/resources/limits.htm

YearCode Name

BrandName Launch

XeonFreq

Cores/

CPU L1 CacheL2

CacheL3

Cache FSB Chipset

SPECint_rate*_

Base2000

SPECfp_rate*_

Base2000TPC-C*tpmC

SAP-SD*

UsersLinpack GFlops

2001 Tualatin Intel Pentium III 1.40Ghz 512M L2 WW32 1.26 1 16K(I)+16K(D) 512KB 0 133 SW HE-SL 12.2 6.5 17079 135 -

2002 Prestonia Intel Xeon 2.20 GHz 512M L2 April 2.20 1 12K(I)+8K(D) 512KB 0 400 E7500 15.7 13.1 33768 215 5.46

2002 Prestonia Intel Xeon 2.80 GHz 512M L2 Sept 2.80 1 12K(I)+8K(D) 512KB 0 400 E7500 18.9 14.4 39007 245 6.72

2003 Prestonia Intel Xeon 3.06 GHz 512M L2 March 3.06 1 12K(I)+8K(D) 512KB 0 533 E7501 21.5 16.7 44943 290 7.83

2003 Prestonia Intel Xeon 3.06 GHz 1M L3 July 3.06 1 12K(I)+8K(D) 512KB 1MB 533 E7501 26.0 20.4 52468 329 8.16

2003 Prestonia Intel Xeon 3.20 GHz 1M L3 Oct 3.20 1 12K(I)+8K(D) 512KB 1MB 533 E7501 26.9 20.7 54097 340 8.40

2004 Prestonia Intel Xeon 3.20 GHz 2M L3 March 3.20 1 12K(I)+8K(D) 512KB 2MB 533 E7501 30.7 24.0 60364 408 9.79

2004 Nocona Intel Xeon 3.60 GHz 1M L2 June 3.60 1 16K(I)+16K(D) 1MB 0 800 E7520 32.1 25.2 68010 470 11.60

2005 Irwindale 64-bit Intel Xeon 3.60 GHz 2M L2 Feb 3.60 1 12K(I)+16K(D) 2MB 0 800 E7520 39.9 31.6 74298 597 13.00

2005 Paxville DP Dual-Core Intel Xeon 2.80 2x2M L2 Oct 2.80 2 12K(I)+16K(D) 2x2MB 0 800 E7520 59.5 39.4 87791 788 18.10

2006 Dempsey Dual-Core Intel Xeon 5080 May 3.73 2 12K(I)+16K(D) 2x2MB 0 1066 5000p 81.2 61.6 125954 1047 22.10

2006 Woodcrest Dual-Core Intel Xeon 5160 June 3.00 2 32K(I)+32K(D) 4MB 0 1333 5000p 123.0 82.1 169360 1285 38.30

2006 Clovertown Quad-Core Intel Xeon X5355 Nov 2.66 4 32K(I)+32K(D) 2x4MB 0 1333 5000p 200.0 103.0 240737 1870 64.00

2007 Clovertown Quad-Core Intel Xeon X5365 July 3.00 4 32K(I)+32K(D) 2x4MB 0 1333 5000p 214 105 251300 2080 70.00

2007 Harpertown Quad-Core Intel Xeon X5460 Nov 3.16 4 32K(I)+32K(D) 2x6MB 0 1333 5000p 242 131 273666 2449 82

http://www.intel.com/performance/resources/limits.htm

http://www.spec.org/cpu2000/results/res2002q1/cpu2000-20020115-01209.html


http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=102021301

http://www.sap.com/solutions/benchmark/pdf/cert3101.pdf











































“Server Compression ”

Source: Intel January 16, 2008.

Performance comparison using SPECjbb2005 bops (business operations per second) between (2004) 2 socket single core Intel Xeon processor (3.6GHz) and (2008) 2 socket Quad-Core Intel Xeon processor E5450 (3.0GHz, 80W, 1333MHz) measured 8/22/07.

1 Floor space based foot on sq ft reduction.

2 Energy costs based on total solution rack power.

3 Energy Savings based on an electric rate of $0.10/kWh assuming 33% average server load.

4 ROI is calculated based on cost of new servers ($6,264 per server pricing based on HP DL 380G5 (32GB RAM) as of Nov 17, 2007. source www.hp.com) divided by energy savings per year.

Actual performance results and savings may vary depending on configuration.

See next slide for detailed calculations


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. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, visit http://www.intel.com/performance/resources/limits.htm or call (U.S.) 1-800-628-8686 or 1-916-356-3104.


Calculations for Refresh Benefits System Configurations:2 Socket Intel® Xeon® Processor 3.6GHz (Nacona)8 x 1GB DDR2-400 DIMMsSystem Idle Power: 225WSystem Workload Power: 382WSPECjbb2005 bops performance: 40487

2 Socket Intel® Xeon® Processor E5450 (3.0GHz 80W HTN)8 x 2GB 667MHz FBDIMMsSystem Idle Power: 237WSystem Workload Power: 353WSPECjbb2005 bops performance: 297,984

Calculations:Floor space: 240sq. Foot to 60 sq. foot (83% reduction)Annual energy costs: 48kW to 6kW (87% reduction)Yearly Energy Savings: $53k ($61.4k - $8.29k) = $53.11kQty 126 2S Intel® Xeon® Processor 3.6GHz (Nacona) servers, at idle power (225W * 16 hours/day), at full load

(382W * 8 hours/day), for one year at 0.10 kWh = $30.69k * 2 (HVAC factor) = $61.4k electricity cost per year.

Compared toQty 17 2S Intel® Xeon® Processor E5450 (HTN 3.0GHz) servers, at idle power (237W * 16 hours/day), at full load

(353W * 8 hours/day), for one year at 0.10 kWh = $4.16k * w (HVAC factor) = $8.29k electricity cost per year.

ROI: Using HP DL 380G5 with 32GB RAM is $6,264 as of Nov 17, 2007 on hp.comCost of 17 servers is $106.5k divided by $53k/year energy savings, this is 2 year ROI

Software:BEA JRockit* (build P27.4.0-3-86647-1.6.0_02-20070801-1931-windows-x86_64, compiled mode



20* Other names and brands may be claimed as the property of others. Copyright © 2008, Intel Corporation. All rights reserved 20

Quad-Core Intel® Xeon® Processor 7300 Series based ServersPerformance Publications and Intel® ISV Application Enabling Results – Config Details

• vConsolidate on Virtual iron: Comparison to Dual-Core Intel® Xeon® Processor 7100 Series (167% claim) based on vConsolidate benchmark result published on Virtual Iron 4.0 software:

– Dual-Core Intel® Xeon® processor 7140M based platform details: Intel server platform with four Quad-Core Intel® Xeon® processors 7140M 3.4GHz, 800 MHz system bus, 16MB L3 Cache, Intel E8500 chipset, 32GB Memory DDr2 PC2-5300P, 14x Seagate Cheetah* 36GB 15krpm Fibre Channel HDD. Maximum performance achieved using 2 CSUs.

– Quad-Core Intel® Xeon® processor X7350 based platform details: Intel server platform with four Quad-Core Intel® Xeon® processors X7350 2.93GHz, 1066 MHz system bus, 8MB L2 Cache, Clarksboro chipset, 32GB Memory FBD PC2-5300F, 14x Seagate Cheetah 36GB 15krpm Fibre Channel HDD. Maximum performance achieved using 4 CSUs.


i/o for virtualized multi-core servers

Documents