vmfs_rdm_perf.pdf

7/29/2019 vmfs_rdm_perf.pdf

1/11

Performance Study

Copyright 2007 VMware, Inc. All rights reserved. 1

Performance Characteristicsof VMFS and RDMVMware ESX Server 3.0.1

VMwareESXServeroffersthreechoicesformanagingdiskaccessinavirtualmachineVMwareVirtual

MachineFileSystem(VMFS),virtualrawdevicemapping(RDM),andphysicalrawdevicemapping.Itisvery

importantto

understand

the

I/O

characteristics

of

these

disk

access

management

systems

in

order

to

choose

therightaccesstypeforaparticularapplication.Choosingtherightdiskaccessmanagementmethodcanbea

keyfactorinachievinghighsystemperformanceforenterpriseclassapplications.

Thisstudyprovidesperformancecharacterizationofthevariousdiskaccessmanagementmethodssupported

byVMwareESXServer.Thegoalistoprovidedataonperformanceandsystemresourceutilizationatvarious

loadlevelsfordifferenttypesofworkloads.Thisinformationoffersyouanideaofrelativethroughput,I/O

rate,andCPUefficiencyforeachoftheoptionssoyoucanselecttheappropriatediskaccessmethodforyour

application.

Thisstudycoversthefollowingtopics:

TechnologyOverviewonpage 2

ExecutiveSummaryonpage 2

TestEnvironmentonpage 2

PerformanceResultsonpage 5

Conclusiononpage 10

Configurationonpage 10

Resourcesonpage 11


2/11


Performance Characteristics of VMFS and RDM

Technology Overview

VMFSisaspecialhighperformancefilesystemofferedbyVMwaretostoreESXServervirtualmachines.

AvailableaspartofESXServer,itisadistributedfilesystemthatallowsconcurrentaccessbymultiplehosts

tofilesonasharedVMFSvolume.VMFSoffershighI/Ocapabilitiesforvirtualmachines.Itisoptimizedfor

storingandaccessinglargefilessuchasvirtualdisksandthememoryimagesofsuspendedvirtualmachines.

RDMisamappingfileinaVMFSvolumethatactsasaproxyforarawphysicaldevice.TheRDMfilecontains

metadata

used

to

manage

and

redirect

disk

accesses

to

the

physical

device.

This

technique

provides

advantagesofdirectaccesstophysicaldeviceinadditiontosomeoftheadvantagesofavirtualdiskonVMFS

storage.Inbrief,itoffersVMFSmanageabilitywiththerawdeviceaccessrequiredbycertainapplications.

YoucanconfigureRDMintwoways:

VirtualcompatibilitymodeThismodefullyvirtualizesthemappeddevice,whichappearstotheguest

operatingsystemasavirtualdiskfileonaVMFSvolume.VirtualmodeprovidessuchbenefitsofVMFS

asadvancedfilelockingfordataprotectionanduseofsnapshots.

PhysicalcompatibilitymodeThismodeprovidesminimalSCSIvirtualizationofthemappeddevice.

VMkernelpassesallSCSIcommandstothedevice,withoneexception,therebyexposingallthephysical

characteristicsoftheunderlyinghardware.

BothVMFSandRDMprovidesuchdistributedfilesystemfeaturesasfilelocking,permissions,persistent

naming,and

VMotion

capabilities.

VMFS

is

the

preferred

option

for

most

enterprise

applications,

including

databases,ERP,CRM,VMwareConsolidatedBackup,Webservers,andfileservers.AlthoughVMFSis

recommendedformostvirtualdiskstorage,rawdiskaccessisneededinafewcases.RDMisrecommended

forthosecases.SomeofthecommonusesofRDMareinclusterdataandquorumdisksforconfigurations

usingclusteringbetweenvirtualmachinesorbetweenphysicalandvirtualmachinesorforrunningSAN

snapshotorotherlayeredapplicationsinavirtualmachine.

FormoreinformationonVMFSandRDM,refertotheServerConfigurationGuide.Executive Summary

Themainconclusionsthatcanbedrawnfromthetestsdescribedinthisstudyare:

Forrandom

reads

and

writes,

VMFS

and

RDM

yield

similar

I/O

performance.

Forsequentialreadsandwrites,RDMprovidesslightlybetterI/OperformanceatsmallerI/Oblocksizes,

butVMFSandRDMprovidesimilarI/OperformanceforlargerI/Oblocksizes(greaterthan32KB).

CPUcyclesrequiredperI/OoperationforbothVMFSandRDMaresimilarforrandomreadsandwrites

atsmallerI/Oblocksizes.AstheI/Oblocksizeincreases,CPUcyclesperI/OoperationincreasesforVMFS

comparedtoRDM.

Forsequentialreadsandwrites,RDMrequiresrelativelyfewerCPUcyclesperI/Ooperation.

Test Environment

Thetestsdescribedinthisstudycharacterizetheperformanceofvariousoptionsfordiskaccessmanagement

VMwareoffers

with

ESX

Server.

We

ran

the

tests

with

auniprocessor

virtual

machine

using

Windows

Server

2003EnterpriseEditionwithSP2astheguestoperatingsystem.ThevirtualmachineranonanESXServer

systeminstalledonalocalSCSIdisk.Weattachedtwodiskstothevirtualmachineonevirtualdiskforthe

operatingsystemandaseparatetestdisk,whichwasthetargetfortheI/Ooperations.Weusedalogical

volumecreatedonfivelocalSCSIdisksconfiguredasRAID0asthetestdisk.WegeneratedI/Oloadusing

Iometer,averypopulartoolforevaluatingI/Operformance(seeResourcesonpage 11foralinktomore

information).SeeConfigurationonpage 10foradetailedlistofthehardwareandsoftwareconfiguration

weusedforthetests.


3/11



Disk Layout

Inthisstudy,disklayoutreferstotheconfiguration,location,andtypeofdiskusedfortests.Weusedasingle

serverwithsixphysicalSASdisksandaRAIDcontrollerforthetests.

Wecreatedalogicaldriveonasinglephysicaldisk.WeinstalledESXServeronthisdrive.Onthesamedrive,

wealsocreatedaVMFSpartitionandusedittostorethevirtualdiskonwhichweinstalledtheWindows

Server2003guestoperatingsystem.

Weconfigured

the

remaining

five

physical

disks

as

RAID

0and

created

a10GB

logical

volume,

referred

to

here

asthetestdisk.WeusedthistestdiskonlyfortheI/Ostresstest.Wecreatedavirtualdiskonthetestdiskand

attachedthatvirtualdisktotheWindowsvirtualmachine.Totheguestoperatingsystem,thevirtualdisk

appearsasaphysicaldrive.

IntheVMFStests,weimplementedthevirtualdisk(seenasaphysicaldiskbytheguestoperatingsystem)as

a.vmdkfilestoredonaVMFSpartitioncreatedonthetestdisk.

IntheRDMtests,wecreatedanRDMfileontheVMFSvolume(thevolumethatheldthevirtualmachine

configurationfilesandthevirtualdiskwhereweinstalledtheguestoperatingsystem)andmappedtheRDM

filetothetestdisk.WeconfiguredthetestvirtualdisksoitwasconnectedtoanLSISCSIhostbusadapter.

Figure 1. Disk layout for VMFS and RDM tests

Fromtheperspectiveoftheguestoperatingsystem,thetestdiskswererawdiskswithnopartitionorfile

system(suchasNTFS)createdonthem.Iometercanreadandwriteraw,unformatteddisksdirectly.Weused

thiscapabilitysowecouldcomparetheperformanceoftheunderlyingstorageimplementationwithout

involvinganyoperatingsystemfilesystem.

ESX Server

.vmdk

Test diskVMFS volume

Operating system diskVMFS volume

Virtual

machine

.vmdk

ESX Server

.vmdk

Test diskVMFS volume


Virtual

machine

.vmdk

ESX Server

Test diskRaw device


Virtual

machine

RDMmapping

file pointerRaw device

mapping

VMFS test RDM test

.vmdk


4/11



Software Configuration

WeconfiguredtheguestoperatingsystemtousetheLSILogicSCSIdriver.OnVMFSvolumes,wecreated

virtualdiskswiththethickoption.Thisoptionprovidesthebestperformingdiskallocationschemefora

virtualdisk.Allthespaceallocatedduringdiskcreationisavailablefortheguestoperatingsystem

immediatelyafterthecreation.Anyolddatathatmightbepresentontheallocatedspaceisnotzeroedout

duringvirtualmachinewriteoperations.

Unless

stated

otherwise,

we

left

all

ESX

Server

and

guest

operating

system

parameters

at

their

default

settings.

I/O Workload Characteristics

EnterpriseapplicationstypicallygenerateI/Owithmixedaccesspatterns.Thesizeofdatablockstransferred

betweentheserverhostingtheapplicationandthestoragealsochanges.Designinganappropriatediskand

filesystemlayoutisveryimportanttoachieveoptimumperformanceforagivenworkload.

Afewapplicationshaveasingleaccesspattern.Oneexampleisbackupanditspatternofsequentialreads.

Onlinetransactionprocessing(OLTP)databaseaccess,ontheotherhand,ishighlyrandom.

Thenatureoftheapplicationalsoaffectsthesizeofdatablockstransferred.Often,thedatablocksizeisnota

singlevaluebutarange.ForMicrosoftExchange,theI/Ooperationsaregenerallysmallfrom4KBto16KB.

OLTPdatabaseaccessesusingMicrosoftSQLServerandOracledatabasesuseamixofrandom8KBreadand

writeoperations.

TheI/Ocharacteristicsofaworkloadcanbedefinedintermsoftheratioofreadtowriteoperations,theratio

ofsequentialtorandomI/Oaccess,andthedatatransfersize.Arangeofdatatransfersizescanalsobe

mentionedinsteadofasinglevalue.

Test Cases

Inthisstudy,wecharacterizetheperformanceofVMFSandRDMforarangeofdatatransfersizesacross

variousaccess

patterns.

The

data

transfer

sizes

we

selected

were

4KB,

8KB,

16KB,

32KB,

and

64KB.

The

access

patternswechosewererandomreads,randomwrites,sequentialreads,sequentialwrites,oramixofrandom

readsandwrites.ThetestcasesaresummarizedinTable1.

NOTE ThethickoptionisnotthedefaultwhenyoucreateavirtualdiskusingtheVIClient.Tocreatevirtualdiskswiththethickoption,weusedacommandlineprogram(vmkfstools).Fordetailsonusing

vmkfstools,seetheVMwareInfrastructure3ServerConfigurationGuide.

Table 1. Test cases

100% Sequential 100% Random

100% Read 4KB,8KB,16KB,32KB,64KB 4KB,8KB,16KB,32KB,64KB

100% Write 4KB,8KB,16KB,32KB,64KB 4KB,8KB,16KB,32KB,64KB

50% Read + 50% Write 4KB,8KB,16KB,32KB,64KB


5/11



Load Generation

WeusedtheIometerbenchmarkingtool,originallydevelopedatIntelandwidelyusedinI/Osubsystem

performancetesting,togenerateI/OloadandmeasuretheI/Operformance.Foralinktomoreinformation,

seeResourcesonpage 11.IometerprovidesoptionstocreateandexecuteawelldesignedsetofI/O

workloads.Becausewedesignedourteststocharacterizetherelativeperformanceofvirtualdisksonraw

devicesandVMFS,weusedonlybasicloademulationfeaturesinthetests.

Iometer

configuration

options

used

as

variables

in

the

tests: Transferrequestsizes:4KB,8KB,16KB,32KB,and64KB.

Percentrandomorsequentialdistribution:foreachtransferrequestsize,weselected0percentrandom

access(equivalentto100percentsequentialaccess)and100percentrandomaccesses.

Percentreadorwritedistribution:foreachtransferrequestsize,weselected0percentreadaccess

(equivalentto100percentwriteaccess),50percentreadaccess(onlyforrandomaccess)and100percent

readaccesses.

Iometerparametersconstantforalltestcases:

NumberofoutstandingI/Ooperations:8

Runtime:5minutes

Rampuptime:60seconds

Numberofworkerstospawnautomatically:1

Performance Results

Thissectionpresentsdataandanalysisofstoragesubsystemperformanceinauniprocessorvirtualmachine.

Metrics

ThemetricsweusedtocomparetheperformanceofVMFSandRDMareI/Orate(measuredasnumberofI/O

operationspersecond),throughputrate(measuredasMBpersecond),andCPUefficiency.

Inthis

study,

we

report

the

I/O

rate

and

throughput

rate

as

measured

by

Iometer.

We

use

an

efficiency

metric

calledMHzperI/OpstocomparetheefficienciesofVMFSandRDM.ThismetricisdefinedastheCPUcost

(inprocessorcycles)perunitI/Oandiscalculatedasfollows:

WecollectedI/OandCPUutilizationstatisticsfromIometerandesxtopasfollows:

IometercollectedI/OoperationspersecondandthroughputinMBps

esxtopcollectedtheaverageCPUutilizationofphysicalCPUs

ForlinkstoadditionalinformationonhowtocollectI/OstatisticsusingIometerandhowtocollectCPU

statisticsusing

esxtop,

see

Resources

on

page 11.

MHz per I/OpsAverage CPU utilization CPU rating in MHz Number of cores

Number of I/O operations per second---------------------------------------------------------------------------------------------------------------------------------------------------------------=


6/11



Performance

Thissectioncomparestheperformancecharacteristicsofeachtypeofdiskaccessmanagement.Themetrics

usedareI/Orate,throughput,andCPUefficiency.

Random Workload

Mostoftheenterpriseapplicationsthatgeneraterandomworkloadstypicallytransfersmallamountsofdata

in

each

I/O

operation.

For

a

small

I/O

data

transfer,

the

time

it

takes

for

the

disk

head

to

physically

locate

the

dataonthediskandmovetowardsitismuchgreaterthanthetimetoactuallyreadorwritethedata.Any

delaycausedbytheoverheadintheapplicationandfilesystemlayerbecomesnegligiblecomparedtotheseek

delay.Inourtestsforrandomworkloads,VMFSandRDMproducedsimilarI/Operformanceasevidentfrom

Figure2,Figure3,andFigure4.

Figure 2. Random mixed /IO per second (higher is better)

Figure 3. Random read I/O operations per second (higher is better)

0

200

400

600

800

1000

1200

1400

1600

4 8 16 32 64

Data s ize (KB)

I/Oo

perationspe

rsecond

VMFS

RDM (Virtual)

RDM (Physical)

0

200

400

600

800

1000

1200

4 8 16 32 64

Data size (KB)

I/Oo

perationspersecond

VMFS

RDM (Virtual)

RDM (Physical)


7/11



Figure 4. Random write I/O operations per second (higher is better)

Sequential Workload

Forsequentialworkloads,applicationstypicallyaccessconsecutivedatablocks.Thediskseektimeisreduced

becausethediskheadhastomoveverylittletoaccessthenextblock.AnyI/Oqueuingdelaythatoccursinthe

applicationandfilesystemlayercanaffecttheoverallthroughput.Accessingfilemetadataanddatastructures

toresolvefilenamestophysicaldatablocksondisksaddsaslightdelaytotheoveralltimerequiredto

completetheI/Ooperation.InordertosaturatetheI/Obandwidth,whichtypicallyisafewhundred

megabytespersecond,applicationshavetoissuemoreI/OrequestsiftheI/Oblocksizeissmall.Asthe

numberofI/Orequestsincreases,thedelaycausedbyfilesystemoverheadincreases.Thusrawdisksyield

slightlyhigherthroughputcomparedtofilesystemsatsmallerdatatransfersizes.Inthesequentialworkload

testswithasmallI/Oblocksize,RDMprovided3to10percenthigherI/Othroughput,asshowninFigure5

andFigure6.

However,asI/Oblocksizeincreases,thenumberofI/OrequestsdecreasesbecausetheI/Obandwidthis

saturated,

and

the

delay

caused

by

file

system

overhead

becomes

less

significant.

In

this

case,

both

RDM

and

VMFSyieldsimilarperformance.

Figure 5. Sequential read I/O operations per second (higher is better)

0

500

1000

1500

2000

2500

3000

4 8 16 32 64Data s ize (KB)

I/Oo

perationspersecond

VMFS

RDM (Virtual)

RDM (Physical)

0

5000

10000

15000

20000

25000

30000

4 8 16 32 64

Data size (KB)

I/Oo

pe

rationspersecond

VMFS

RDM (Virtual)

RDM (Physical)


8/11



Figure 6. Sequential write I/O operations per second (higher is better)

Table2andTable3showthethroughputratesinmegabytespersecondcorrespondingtotheaboveI/O

operations

per

second

for

VMFS

and

RDM.

The

throughput

rates

(I/O

operations

per

second

*

data

size)

are

consistentwiththeI/OratesshownaboveanddisplaybehaviorsimilartothatexplainedfortheI/Orates.

CPU Efficiency

CPUefficiencycanbecomputedintermsofCPUcyclesrequiredperunitofI/Oorunitofthroughput(byte).

WeobtainedafigureforCPUcyclesusedbythevirtualmachineformanagingtheworkload,includingthe

virtualizationoverhead,bymultiplyingtheaverageCPUutilizationofalltheprocessorsseenbyESXServer,

theCPUratinginMHz,andthetotalnumberofcoresinthesystem(fourinthiscase).Inthisstudy,we

measuredCPUefficiencyasCPUcyclesperunitofI/Ooperationspersecond.

CPUefficiencyforvariousworkloadsisshowninfigures7through11.Forrandomworkloadswithsmaller

I/Oblocksizes(4KBand8KB)CPUefficiencyobservedwithbothVMFSandRDMisverysimilar.ForI/Oblock

sizesgreaterthan8KB,RDMofferssomeimprovement(7to12percent).Forsequentialworkloads,RDMoffers

an8to12percentimprovement.Aswithanyfilesystem,VMFSmaintainsdatastructuresthatmapfilenames

Table 2. Throughput rates for random workloads in megabytes per second

Random Mix Random Read Random Write

Data Size (KB) VMFS RDM (V) RDM (P) VMFS RDM (V) RDM (P) VMFS RDM (V) RDM (P)

4 5.54 5.46 5.45 4.2 4.14 4.15 9.78 9.7 9.64

8 10.39 10.28 10.27 8.03 7.9 7.9 17.95 17.95 17.92

16 18.63 18.43 18.43 14.49 14.3 14.34 31.58 31.2 31.06

32 31.04 30.65 30.65 24.7 24.39 24.38 50 49.24 49.45

64 46.6 45.8 45.78 38.11 37.69 37.66 71 70.15 70.01

Table 3. Throughput rates for sequential workloads in megabytes per second

Sequential Read Sequential Write

Data Size (KB) VMFS RDM (V) RDM (P) VMFS RDM (V) RDM (P)

4 87.53 90.66 95.17 106.14 107.47 109.56

8 161.04 168.47 177.44 172.38 173.81 174.86

16 300.44 311.53 316.92 221.8 226.21 229.05

32 435.92 437.94 437.47 252.74 265.89 271.2

64 436.88 438.04 438.06 308.3 315.63 317

0

5000

10000

15000

20000

25000

30000

4 8 16 32 64

Data size (KB)

I/Oo

peratio

nspersecond

VMFS

RDM (Virtual)

RDM (Physical)


9/11



tophysicalblocksonthedisk.EachfileI/Orequiresaccessingthemetadatatoresolvefilenamestoactualdata

blocksbeforereadingdatafromorwritingdatatoafile.ThenameresolutionrequiresafewextraCPUcycles

everytimethereisanI/Oaccess.Inaddition,maintainingthemetadataalsorequiresadditionalCPUcycles.

RDMdoesnotrequireanyunderlyingfilesystemtomanageitsdata.Dataisaccesseddirectlyfromthedisk,

withoutanyfilesystemoverhead,resultinginalowerCPUcycleconsumptionandhigherCPUefficiency.

Figure 7. CPU efficiency for random mix (lower is better)

Figure 8. CPU efficiency for random read (lower is better)

Figure 9. CPU efficiency for random write (lower is better)

Figure 10. CPU efficiency for sequential read (lower is better)

0.00

0.20

0.40

0.60

0.80

1.00

4 8 16 32 64

Data s ize (KB)

Efficiency(MHzperI/Ops

)

VMFS

RDM (Virtual)

RDM (Physical)

0.00

0.20

0.40

0.60

0.80

1.00

4 8 16 32 64

Data size (KB)

Efficiency(MHzperI/Op

s)

VMFS

RDM (Virtual)

RDM (Physical)

0.00

0.20

0.40

0.60

0.801.00

4 8 16 32 64Data s ize (KB)

Efficiency(MHzperI/Op

s)

VMFS

RDM (Virtual)

RDM (Physical)

0.00

0.20

0.40

0.60

0.80

1.00

4 8 16 32 64Data s ize (KB)

Efficiency(MHzperI/

Ops)

VMFS

RDM (Virtual)

RDM (Physical)


10/11



Figure 11. CPU efficiency for sequential write (lower is better)

Conclusion

VMwareESXServeroffersthreeoptionsfordiskaccessmanagementVMFS,RDM(Virtual),andRDM

(Physical).Alltheoptionsprovidedistributedfilesystemfeatureslikeuserfriendlypersistentnames,

distributedfilelocking,andfilepermissions.BothVMFSandRDMallowyoutomigrateavirtualmachine

usingVMotion.Thisstudycomparesthethreeoptionsandfindssimilarperformanceforallthree.

Forrandom

workloads,

VMFS

and

RDM

produce

similar

I/O

throughput.

For

sequential

workloads

with

smallI/Oblocksizes,RDMprovidesasmallincreaseinthroughputcomparedtoVMFS.However,the

performancegapdecreasesastheI/Oblocksizeincreases.VMFSandRDMoperateatsimilarCPUefficiency

levelsforrandomworkloadsatsmallerI/Oblocksizes.Forsequentialworkloads,RDMshowsimprovedCPU

efficiency.

TheperformancetestsdescribedinthisstudyshowthatVMFSandRDMprovidesimilarI/Othroughputfor

mostoftheworkloadswetested.Mostenterpriseapplications,suchasWebserver,fileserver,databases,ERP,

andCRM,exhibitI/Ocharacteristicssimilartotheworkloadsweusedinourtestsandhencecanuseeither

VMFSorRDMforconfiguringvirtualdiskswhenruninavirtualmachine.However,therearefewcasesthat

requireuseofrawdisks.BackupapplicationsthatusesuchinherentSANfeaturesassnapshotsorclustering

applications(forbothdataandquorumdisks)requirerawdisks.RDMisrecommendedforthesecases.We

recommenduseofRDMforthesecasesnotforperformancereasonsbutbecausetheseapplicationsrequire

lowerlevel

disk

control.

Configuration

Thissectiondescribesthehardwareandsoftwareconfigurationsweusedinthetestsdescribedinthisstudy.

Server Hardware

Processors:2dualcoreIntelXeonprocessor5160,3.00GHz,4MBL2cache(4corestotal)

Memory:8GB

Localdisks:

1Seagate146GB10KRPMSCSI(forESXServerandtheguestoperatingsystem)

5Seagate146GB10KRPMSCSIinRAID0configuration(forthetestdisk)

Software

Virtualizationsoftware:ESXServer3.0.1(build32039)

Guest Operating System Configuration

Operatingsystem:WindowsServer2003R2EnterpriseEdition32bit,ServicePack2,512MBofRAM,1

CPU

Testdisk:10GBunformatteddisk

0.00

0.20

0.40

0.60

0.80

1.00

4 8 16 32 64Data size (KB)

Efficien

cy(MHzperI/Ops)

VMFS

RDM (Virtual)

RDM (Physical)


11/11

11


VMware, Inc. 3401 Hillview Ave., Palo Alto, CA 94304 www.vmware.com

Copyright 2007 VMware, Inc. All rights reserved. Protected by one or more of U.S. Patent Nos. 6,397,242, 6,496,847, 6,704,925, 6,711,672, 6,725,289, 6,735,601, 6,785,886,6,789,156, 6,795,966, 6,880,022, 6,944,699, 6,961,806, 6,961,941, 7,069,413, 7,082,598, 7,089,377, 7,111,086, 7,111,145, 7,117,481, 7,149, 843, 7,155,558, 7,222,221, 7,260,815,7,260,820, 7,269,683, 7,275,136, 7,277,998, 7,277,999, 7,278,030, 7,281,102, and 7,290,253; patents pending. VMware, the VMware boxes logo and design, Virtual SMP andVMotion are registered trademarks or trademarks of VMware, Inc. in the United States and/or other jurisdictions. Microsoft, Windows and Windows NT are registeredtrademarks of Microsoft Corporation. Linux is a registered trademark of Linus Torvalds. All other marks and names mentioned herein may be trademarks of their respectivecompanies.Revision 20071207 Item: PS-035-PRD-01-01

Iometer Configuration

NumberofoutstandingI/Os:8

Rampuptime:60seconds

Runtime:5minutes

Numberofworkers(threads):1

Accesspatterns:

random/mix,

random/read,

random/write,

sequential/read,

sequential/write

Transferrequestsizes:4KB,8KB,16KB,32KB,64KB

Resources

ToobtainIometer,goto

http://www.iometer.org/

FormoreinformationonhowtogatherI/OstatisticsusingIometer,seetheIometerusersguideat

http://www.iometer.org/doc/documents.html

TolearnmoreabouthowtocollectCPUstatisticsusingesxtop,seethechapterUsingtheesxtopUtility

intheVMwareInfrastructure3ResourceManagementGuideathttp://www.vmware.com/pdf/vi3_301_201_resource_mgmt.pdf

ForadetaileddescriptionofVMFSandRDMandhowtoconfigurethem,seechapters5and8ofthe

VMwareInfrastructure3ServerConfigurationGuideathttp://www.vmware.com/pdf/vi3_301_201_server_config.pdf

VMwareInfrastructure3ServerConfigurationGuidehttp://www.vmware.com/pdf/vi3_301_201_server_config.pdf

vmfs_rdm_perf.pdf

Documents