RAID 6 or RAID 10 - Spiceworks Community

文章推薦指數: 80 %
投票人數:10人

RAID 10 is faster and safer. But RAID 6 is pretty safe on its own (your array is not crazy large.) So both are completely good options. If you ... Home News&Insights SpiceworksOriginals Snap! Features Roundups Polls VoiceofIT(VoIT) Videos Podcasts Community Askquestion CommunityHome Cloud Collaboration Networking WaterCooler Windows Allforums How-Tos Scripts Vendors Meetups Reviews OnlineEvents Login Join Home Storage&Backup DataStorage RAID6orRAID10 Postedbymichaelprivitera Solved DataStorage Helloall, I'vebeendoingalotofreadinglatelyaboutRAID6andRAID10andhavetodecidequicklyonaservermycompanyisputtinginplayshortly. Here'sthescenario: MycompanypurchasedanewDellR720foraclientwithdualProcsand128GBofRAM. Wealsohave6x2TBSATA7.2kRPMdrivesinthere. We'relookingtoputabout5VMs(DCs,terminalservers,fileservers,etc.,NOExchange)onthisandleavethehostmachinealoneassimplyaHyper-Vserver(Server2012R2). IconfiguredtheRAIDasonelargeRAID6volume(~7.4TBofusablestorage)andpartitionedwithinWindowsas1x300GBpartitionand1x~7TBpartition. Isthissetupoptimal? OrshouldIbethinkingtogoRAID10? IknowgoingRAID10meansa2TBlossofusablestorage,butIthinkIneedtoworrymoreaboutperformanceinasetuplikethis. Anythoughts? Andthanksinadvance! Spice(27) Reply(80) flagReport michaelprivitera sonora PopularTopicsinDataStorage AremechanicalNASsafemountedaboveorbelowrackmountups? Harddriveshowingspace,butwontmovedataasitreturnsanospa... BuildingabackuptargetforVeeam-WhichRAIDcontroller harddrivelockedbyhiberfil.syscopydata iscsitraficseparationVLANvsphysicalport Viewalltopics check BestAnswer ScottAlanMiller purecapsaicin checkBestAnswer RAID10isfasterandsafer. ButRAID6isprettysafeonitsown(yourarrayisnotcrazylarge.) Sobotharecompletelygoodoptions. Ifyouneedthecapacity,thenRAID6isthechoice. Ifyoudon'tneedthecapacity,RAID10issuperiorineverywaytoRAID6. Soitreallycomesdowntothecapacityquestion. Isthatcapacityimportantornot. Decidethatandyouknowyouranswer. Spice(28) flagReport 1foundthishelpful thumb_up thumb_down ViewBestAnswerinrepliesbelow 80Replies Prev12Next ScottAlanMiller purecapsaicin checkBestAnswer RAID10isfasterandsafer. ButRAID6isprettysafeonitsown(yourarrayisnotcrazylarge.) Sobotharecompletelygoodoptions. Ifyouneedthecapacity,thenRAID6isthechoice. Ifyoudon'tneedthecapacity,RAID10issuperiorineverywaytoRAID6. Soitreallycomesdowntothecapacityquestion. Isthatcapacityimportantornot. Decidethatandyouknowyouranswer. Spice(28) flagReport 1foundthishelpful thumb_up thumb_down stables2511 habanero IthinkiamrightinsayingforRAID10youneed4or8drives?sodoingiton6wouldleaveyouwith2redundantfailoverdrives. ForyourpurposesiwouldgoRAID6asthemachineiswellspecedandperformanceshouldntbeanissue. Spice(1) flagReport 0of2foundthishelpful thumb_up thumb_down NerdHerder TheNerdHerdersisanITserviceprovider. anaheim IwoulduseRAID10. Yourdrivesare7200andSATAnotSASor10K. AssumingyourVMwillbesomesortofserversystem(SQL,MAIL,etc)youmaywanttheperformance. flagReport Wasthisposthelpful? thumb_up thumb_down _Justin_ ghostchili stables2511wrote: IthinkiamrightinsayingforRAID10youneed4or8drives?sodoingiton6wouldleaveyouwith2redundantfailoverdrives. ForyourpurposesiwouldgoRAID6asthemachineiswellspecedandperformanceshouldntbeanissue. Raid10canbedonewith6drives.Aslongasyouhaveanevendrivecountandatleast4drives,youcandoRAID10.It'sastripeofmirrors,sothinkofitasaRAID0acrossabunchofRAID1mirrors.With2disksyouwouldjusthaveRAID1.With4youwouldhaveastripeacrosstwomirrors.With6disksyouhaveastripeacrossthreemirrors,etc. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down NazihHaddad Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. CedarsHostisanITserviceprovider. thaipepper that'sreallynicebutwhySATA? Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down OP michaelprivitera sonora @Nazihitwasacostsituation. Wehadtocutcornersinplacestogetthemtosign. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down OP michaelprivitera sonora IsRAID10expandable? OrifmorestorageisneededIwouldjustcreateanotherRAIDarrayif10isnot? Also,canyouconfigureahotsparewithRAID10? flagReport Wasthisposthelpful? thumb_up thumb_down NazihHaddad Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. CedarsHostisanITserviceprovider. thaipepper forraid10hotspare.checkthis http://community.spiceworks.com/topic/197987-raid-10-do-i-need-hot-spares inregardstotheexpansionyoucandothataslongasyourraidcontrollersupportitandyoucanalwaysaddmoredrivesandcreateanewraid10array. flagReport Wasthisposthelpful? thumb_up thumb_down Brianinca Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. ghostchili SkipthehotsparewithRAID10,getacoldspareandkeepitontheshelf.Don'tdoRAID6unlessyouhavearealsmartcontrollerwithabigcache.Givenyourbudgetconstraintsyoudon'thaveone,theLSIcontrollersthatDellusesaregenerallyprettybasic.SetupnotificationandTESTthatyou'llgetmessagedifadrivedrops,butuseRAID10.  YoureallyhaveatonofRAMinthatboxgivenhowmanyVM'syouplantohost.Doyouhaveplansforadatabaseserverorsomething? Regards, BrianinCA Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down OP michaelprivitera sonora Onemorequestion,thestripsize. IneitherRAID6or10,isthedefault64kbok,orshouldIgoupto128kb? flagReport Wasthisposthelpful? thumb_up thumb_down OP michaelprivitera sonora @Briannowedon't,wejustwentoverkillontheRAM. AfterdoingallofthereadingI'vedoneonthiscommunity(whichisanabsolutelystellarcommunitybytheway)Iprobablywouldhavere-thoughtsomeaspectsofourapproachwiththisbox. Ifyou'renotlearningyou'renotlivingthough,right?lol. Spice(5) flagReport Wasthisposthelpful? thumb_up thumb_down feliperaffo pimiento MypreferenceisRaid10.playitsafe! flagReport Wasthisposthelpful? thumb_up thumb_down Chris(EMC) anaheim +1fortheRAID10plusColdSpare. flagReport Wasthisposthelpful? thumb_up thumb_down Goalieman tabasco IlikeRAID10fortherebuildtime.It'smuchfasterthanRAID6.Helpsonperformance. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down webby chipotle michaelpriviterawrote: Onemorequestion,thestripsize. IneitherRAID6or10,isthedefault64kbok,orshouldIgoupto128kb? Dependsonthesizeofyourfiles.ThespeedbenefitofRAID10(AlsoRAID0,5and6)comesfromspreadingthefileovermultipledisks,soifyouraveragefilesizeissmallerthan:StripeSize*numberofusabledisksYouarewastingthisbenefit(asthedatawillonlybespreadacrossoneortwodisks)soyouneedasmallerstripetogetmaximumthroughput.However,ifyouhavealargeaveragefilesize:Filesize/(StripeSize*numberofusabledisks)=numberofreadoperationstoretrievethefileInthiscasea64KBstripewilltaketwiceasmanyreadoperationsasa128KBstripeandhencebeslower. Spice(3) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin michaelpriviterawrote: Onemorequestion,thestripsize. IneitherRAID6or10,isthedefault64kbok,orshouldIgoupto128kb? Dependsonworkload,notRAIDtype(mostly.) Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down KOOLER(StarWind) Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. BrandRepresentativeforStarWind mace RAID10.RAID6hasaverylimitedamountofusecases(videocapture,clusteredinRAINwithlog-structuredfilesystemcombinedetc).  flagReport Wasthisposthelpful? thumb_up thumb_down DeonCui jalapeno InmyopiniontherandomwriteIOPSlossfromRAID6isnotworthitvsRAID10inavirtualenvironmentforstorageofbootdisksanddrivescontainingapplications. AlsoinRAID6youhavetocalculateparitytwicewhichdependingonyourcontrollerisnegligibleorabigperformanceloss. ThemarginalcostofaddinganotherpairofdrivestoincreasetheRAID10arraywouldbeeasierthantryingtoincreaseIOperformancelateronaRAID6array. flagReport Wasthisposthelpful? thumb_up thumb_down ScottManning habanero Nice,RAID10forsure. flagReport Wasthisposthelpful? thumb_up thumb_down OliverSchwieger anaheim IfyouhaveorderedtheServerwithoneoftheHighPerformanceRaidControllers,useRaid6asitwilluse4HarddrivesforRead/Write.IfyouconfigureRaid10itwillonlyuse3drivesforPerformance,theother3willjustbethemirrorsandwontadduptored/write/IOPSPerformance.IfyouusethecheapRaidController,youwillhaveverylowwritePerformanceifyouuseRaid6or5. WehavemadesomePerformancechecksonaR720xdwith 20drivesandtheH710PController.Raid6wasalotfasterthanRaid10withthisconfig.DrawbackofRaid6isthatyouwillhaveawaylongerrebuildtimeifyouhaveafaultydrive.Performancewillbeabitslowertoointhatcase. IwouldntrecommendtouseSATAdrivesinaVMEnivronment.Theyarewaytoslow.BetterwouldbetousesomeSAS10kdrives flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: IfyouhaveorderedtheServerwithoneoftheHighPerformanceRaidControllers,useRaid6asitwilluse4HarddrivesforRead/Write.IfyouconfigureRaid10itwillonlyuse3drivesforPerformance,theother3willjustbethemirrorsandwontadduptored/write/IOPSPerformance. Sorry,that'snothowRAIDworks. RAID10blowsRAID6performanceoutofthewater,everytime. Whydoyoufeelthathalfofthemirrorsitsidle?BothRAID6andRAID10useeverydriveforreadperformance. RAID6hasa6xpenaltyonwrites,RAID10hasa2x. SotheyareequalinreadsandRAID10is300%fasteronwrites. Spice(3) flagReport 1foundthishelpful thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: WehavemadesomePerformancechecksonaR720xdwith 20drivesandtheH710PController.Raid6wasalotfasterthanRaid10withthisconfig. Somethingwasseriouslywrongwiththesetup. It'sphysicallyimpossibleforequalsizedRAID6tokeepupwithRAID10.  Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down OliverSchwieger anaheim Raid10Setupusesonly10drivesforRead/Write/IOPS.Theother10drivesarejustmirrorsoftheother10.IbelievedthatRaid10isfasterforyears,butthisdoesntseemtobetrueanymorewiththenewGenerationsofRaidControllersanymore.CPUsonthosearesofastthatparitycalculationsdoesntlowertheOverallPerformanceanymore.DellrecommendedRaid5/6SetupevenonthePowervault3220afterwehadseverePerformanceissueswithavirtualizedSQLServer.WedidaonlineconversionofourRaid10configtoRaid5anditwasperformingalotfaster.Intheendwemovedthedatabasetoaraid5on7SSDs+1ashotsparetogettheIOPSandreadPerformanceweneeded. Spice(1) flagReport 0of1foundthishelpful thumb_up thumb_down Brianinca Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. ghostchili SSD'sarenothingatalllikespinnydisksfori/operformance,you'recomparingapplesandoctopuses.YourunderstandingofRAID10isflawed,aswell. Regards, BrianinCA Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: Raid10Setupusesonly10drivesforRead/Write/IOPS.Theother10drivesarejustmirrorsoftheother10. It'snotsupposedtodothat. TechnicallyeachimplementationcandoanythingthattheywantforanyRAIDlevel,RAIDitselfdoesn'tforcespecificread/writepatterns. Buteverydriveshouldbeusedforreads,thereisnoreasonforitnotto. Itshouldnotbejustmirrors,butmirrorsusedinpairs. ThishasbeenthestandardimplementationgoingbacktotheveryfirstRAID10arraysintheearly1990s.Doyouhavedocumentationonaspecificcontrollerthatisnotworking? THisisamajorissuewiththatcontroller.SinceeachpairisaRAID1,whatyouaresayingisthatRAID1isbrokenaswell. Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: CPUsonthosearesofastthatparitycalculationsdoesntlowertheOverallPerformanceanymore. EvenifCPUswereinfinitelyfast,RAID6wouldonlymatchRAID10inreadperformanceandstillbe1/3rdofitswriteperformance.NoCPUistakenintoaccountinthatcalculation. AnyCPUoverheadaddsasmall,additionallatencytoeveryRAID6operationthatdoesnotexistinRAID10...butthatisontopofthe"matching"readspeedand1/3rdwritespeedissues. flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: DellrecommendedRaid5/6SetupevenonthePowervault3220afterwehadseverePerformanceissueswithavirtualizedSQLServer. DellissostronglyopposedtoRAID5,forgoodreason,thattheyremoveditfromtheirGUIinterfacestoprotectpeoplefromselectingitandthinkingthatDellrecommenditsimplybyincludingit. RAID5isnot,inanyway,recommendedbythem. RAID6andRAID10,yes,notRAID5. flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin DellranksRAID10asoutperformingRAID6ontheirPERC6controller. Sothisisn'tanissueonthePERCs. http://en.community.dell.com/techcenter/storage/w/wiki/perc-6-performance-analysis-report.aspx flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin Whendoingdatabaseperformancetesting,DellusesRAID10todemonstratethespeedlimitsoftheircontrollers... http://www.dell.com/downloads/global/products/pvaul/en/perc-improvements-dell-poweredge-12th-generat... flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin DellexplicitlystatesthatRAID10useseverydriveinthearrayforreadperformance: http://www.dell.com/downloads/global/power/ps3q07-20070368-Olivarez-OE.pdf flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OldDellperformancestudyofRAID10outperformingRAID5fordatabases... http://www.dell.com/downloads/global/solutions/tradeoffs_RAID5_RAID10.pdf?c= flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin DellremovingRAID5toprotectitscustomersfrommisunderstandingtheimplicationsofRAID5... http://community.spiceworks.com/topic/252000-sam-won-the-raid-5-battle flagReport Wasthisposthelpful? thumb_up thumb_down NazihHaddad Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. CedarsHostisanITserviceprovider. thaipepper JustwonderingnowwhydoVMwareandsomeothersstillrecommendRaid5asachoice.isitjustforhavinglargerdiskspace? flagReport Wasthisposthelpful? thumb_up thumb_down SWEMans pimiento Ireadthisrecently,butIdon'tunderstandtheformula.AccordingtothisarticleRAID6willbenogoodin2019. http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805 flagReport Wasthisposthelpful? thumb_up thumb_down JohnEclipse anaheim IwouldgoRAID10aswell.IreconfiguredmyRAID6toaRAID10recentlyandnoticedamajorimprovementinperformance.  flagReport Wasthisposthelpful? thumb_up thumb_down OliverSchwieger anaheim WehadconfiguredourDellPowervaultMD3220withaRaid10on18disks(10kSAS900GBdrives)plus2drivesashotspare.Plus4200GBSSDsconfiguredasaCache.OneServerconnectedwithaDualChannelSASControllerwasperformingverybadly(reallifeMS-SQLServerand severalbenchmarks).Iconfiguredeverythingaccordingtomy20yearsofITexperienceandalsoaccordingtowhatDellpre-salesrecommended. Afterwereportedtheperformance problemstoDell,Dell consultantsexplicitlyrecommendedchangingraidLevelto5or6andtonotuseSSDsasCache(SSDsasCacheonlyaddedmorewritePerformanceandnoreadPerformance).wehavearead/writeratioof90/10.SotheSSDswherenearlyusesless.WechangedtheconfigandcouldmeasureaslightincreaseofIOPSandreadPerformancewithraid6on18diskscomparedtoraid10onthesamedisks.ThedatabasewemovedtoaRaid5onSSDs(weaddedaMD1220with6moreSSDs. @ScottWhenyouthorouglylookintothepdfcomparingtheperformanceofRaid10 andRaid5,youseethatifyoudonthitmaximumTransactions,raid5performsnearlyasfastasraid10.StillDellrecommendsRaid10fornearlyeveryconfiguration,evenforfileservers.ThismightbetrueforafileServerHostingDatafor10kusers,butsurelynotforamediumsizedServerHostingdatafor100users.DelljustwantstosellabunchloadofharddrivesandstorageArrays.Ipersonallywouldntspendanother15kEURonourstorageforjust5%morePeakPerformance.AlsoyousaythatRaid5has1/3lesswritePerformancewhichdefinatlyisnttrue,evenaccordingtotheBenchmarksfromDellyoulinkedhere.Itsmorelike5-10%lesswritePerformance.MostoftheBenchmarksifoundutilize20-100disksfortheBenchmark.InaSOHOconfigurationyouusuallyhavealotlessdisks. Spice(4) flagReport Wasthisposthelpful? thumb_up thumb_down steveweltman Newcontributor pimiento HiMichaelPrivitera ThedecisiontorunVMsonasingledualCPUhostassumesalotaboutthecurrentinboardcontrollerssupportcapability.IwouldsuggestyourcompanyconsiderssolvingthefutureI/Oissuesaheadoftimewithanoffsystemstoragearray.YouwillneedaSANandasuitablecontrollersothatyoucanmaximizetheI/OrequiredfortheVMswhichmaybefinewhennotunderuserloadbutwillwhimperwhentheclientstartsrelyingonitsservices.AsolidHBAwithamodestcontrollerandmodestSANwillnotcrushtheirbudgethoweverwillsustaintheclientsdatagrowthtrajectorywellintothelifecycleoftheserver.ForaRAIDservice,itreallydependsontheclientsobjectivesofdataprotection.Forsimplicity,useRAID6.Themorespinningplatters(orbetter,enterprisegradeSSDs)inyourprimaryarray,thebetteryourI/O. Wishingyouthebestinyourengagement, SteveWeltman flagReport Wasthisposthelpful? thumb_up thumb_down PhilipCoakes serrano Haveyouconsideredrebuildperformancewhen(notif)youhaveadrivefailurewithinanarray? RAID6willhaveanoticeablydetrimentalaffectonperformanceduringarebuild,whereasRAID10ismuchkindertothehost&workloads. flagReport Wasthisposthelpful? thumb_up thumb_down BryceKatz Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. mace Brianincawrote: you'recomparingapplesandoctopuses. **cleansMtDewfrommonitor,keyboard,ANDdesk**Wellplayed,Sir.Wellplayed. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down detox1978 sonora Iknowthisisanoldtopic,butthoughtI'daddmythoughtsafterarecentchatwithDell. DellnolongerrecommendRAID10astheirpreferredconfiguration,becausewhenadiskfailsinRAID10itusestheonlyworkingdiskwiththatdatatorebuildthemirrorpair. OriginallyitwasthoughttheonlyriskinRAID10wasthatanyfaultsonoriginalcopyaretransferredtothenew. However,DellnowhighlightthatputtingalotofIOPSontotheonlyworkingcopyofadisktocreateanewmirrorpairisrisky;andtheyhaveseentheenoughfailuresduringtherebuildprocesstochangetheirstance. Spice(3) flagReport Wasthisposthelpful? thumb_up thumb_down OliverSchwieger anaheim well,theydidn'ttellustheexactreason,butyourexplanationseemsreasonable. inthemeantimewechangedtheconfigfromraid5toraid10(10active200GBSSDplus1hotspare)anddidn'tseeanychangeinperformance.Westillconsiderchangingitback,oreventrytochangetheconfigtoraid6. Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down D8805 Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. datil HavingreadquiteafewofthesethreadsacoupleofyearsagothatgaveclearandoverwhelmingsupporttoselectingRAID10overRAID6,Idecided,basedontheperformanceofRAID10,fora4HDDRAID10array. Inowregrethavingdoneso. Whatisdescribedabovebyphilhowell(andDell)isexactlywhathappened.Therewasadrivefailure.Iswappedinareplacementdrive,thenthe"good"sideoftheMirrorstartedgivingerrorsandcouldnotfinishtherebuild.Fark!(isn'tthiswhyIavoidedRAID5?) AtthismomentIamthreedaysintoaDDcloneofthe3TB"gooddrivewitherrors"thatholdsdatanecessaryforasuccessfulrebuild,hopingtoresurrectitandsalvagemylocalbackuprestorepoints. ThereisstillanoffsitecopyofthedataonAmazonS3/Glacier,butIhavefounditisproblematictoretrieveasitisintegratedwiththebackupsoftwarewhichneedsaccesstodataontheRAIDarraytoreferenceforretrieval.DoubleFark! Plus,thebackupserverlivesontheVMHostIhavenowre-purposedasaLinux/TrinityRescueKitlifesupportmachineforEpisodeIIofAttackoftheClones.Ourresourcesaremeager. IthinkI'lltaketheperformancehitwhenthisisresolvedandconverttoRAID6,asIdon'twanttodothisdanceagain. ForgoingonthreeweeksnowI'vebeensweatingbullets,fallingbackonWindowsBackupformySBSDC/Exchange/FileServerandcopyingmyVantagedatabasebackupmanuallyeachdayasIisolatedtheproblemandworktowardresolution. Inthefutureitmightbeprudenttoalsocopythecontentsofthearraytoaremovablediskperiodicallyandkeepthatinthedatasafe. IthinkRAID10stilltrumpsRAID5handsdown,buttheadvantagesofRAID6aremoresubtle.ThisRAID10failduringrebuildisessentiallythereasonIavoidedconsideringRAID5,astheoddsofafailurearealmostguaranteedonRAID5withlargemedia. WithRAID6spreadingthelovearoundtoanumberofdrives,evenifthereisaperformancehit,itseemsthiswillstilloffermebetterprotectionfromloss,oncetheaspectofasuccessfulrebuildisbakedintotheequation. ThisISthereasonweuseRAIDinthefirstplace,right? Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin D8805wrote: Whatisdescribedabovebyphilhowell(andDell)isexactlywhathappened.Therewasadrivefailure.Iswappedinareplacementdrive,thenthe"good"sideoftheMirrorstartedgivingerrorsandcouldnotfinishtherebuild.Fark!(isn'tthiswhyIavoidedRAID5?) Sowhydoyouregretit? NeitherRAID5,norRAID6,norRAID10wouldlikelyhaveprotectedyouinthiscase. ButRAID10gaveyouthebestchanceofit. Soareyouregrettingprotectingyourselfasbestaspossible? Nothingis100%,nothing. Ever. Shortofthat,whywouldyouregrethavingprotectingyourselfthebestthatyoucould? flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin philhowellwrote: Iknowthisisanoldtopic,butthoughtI'daddmythoughtsafterarecentchatwithDell. DellnolongerrecommendRAID10astheirpreferredconfiguration,becausewhenadiskfailsinRAID10itusestheonlyworkingdiskwiththatdatatorebuildthemirrorpair. OriginallyitwasthoughttheonlyriskinRAID10wasthatanyfaultsonoriginalcopyaretransferredtothenew. However,DellnowhighlightthatputtingalotofIOPSontotheonlyworkingcopyofadisktocreateanewmirrorpairisrisky;andtheyhaveseentheenoughfailuresduringtherebuildprocesstochangetheirstance. Doyouhaveareferenceforthis? Thissoundslikehorriblyrecklessadvice. Whatareyouthinkingthattheyrecommendnow?TheriskthatyouaresayingthattheysayexistswithRAID10mightexist,butitiswellknowntoexistwiththealternativesacrossALLdrives,whichisworse. SothelogicthatyouareusinghereastowhytoavoidRAID10,wouldactuallytellyoutoUSERAID10. BasicallyRAIDisn'tperfectandhasrisks. Yup,that'showthingswork. ButRAID10remainsthesafestoptionbyfar. flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin D8805wrote: IthinkRAID10stilltrumpsRAID5handsdown,buttheadvantagesofRAID6aremoresubtle. As,apparently,aretherisks. RAID10issaferthanRAID6,byabit. Inonlyfourdrives,theyareprettycloseandRAID10mostlywinsonperformance. ButthesafetyofRAID10remains. Thatyoumighthavefoundanedgecasedoesn'timplythatyoudidsomethingrisky. Youareusingananecdotetodeterminearate,whichyoushouldneverdo. Thisistheoppositeofthenormal"drivingwithnoseatbelt"problem. Inthenormalproblem,the"Igotlucky"scenario,peoplepointouthowoftentheydosomethingrisky,likeuseRAID5,andgetawaywithit. Sameasspeedingandwearingnoseatbelt. Onaverageyoucandothatalotbeforeyoudie,butnoonethinksthatdrivingwithoutaseatbeltissmart.Thisistheopposite. Youhavethecasewheresomeonewasdrivingsafely,woreaseatbeltandsomethingfailedanditwastheseatbeltthatkilledhim. Yup,thathappens.Butitdoesn'thappenoften. Justbecauseyouknewtheguythishappenedto(diedfromwearingaseatbelt)thesafedecisionisnottoavoidseatbeltsbuttokeepwearingthem. Because,onaverage,seatbeltsprotectyou. SameasRAID10overRAID6. Onaverage,RAID10savesyourdatamorethanRAID6. ButthatdoesnotimplythatRAID10ininfallible,onlysafer. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin D8805wrote: WithRAID6spreadingthelovearoundtoanumberofdrives,evenifthereisaperformancehit,itseemsthiswillstilloffermebetterprotectionfromloss,oncetheaspectofasuccessfulrebuildisbakedintotheequation. ThisISthereasonweuseRAIDinthefirstplace,right? RAID6doesnotofferbetterprotectionfromloss. Themathjustdoesn'taddupthere. Especiallyasyourdrivesgetbiggerandthespindlecountincreases. Thebiggerthedrives,theheaviertheutilization,themorespindles,thelargerthearrayallmakeRAID10moreandmoresafeincomparisontoRAID6.RAID6hassomanyfailuremodesthatRAID10doesnot,Idon'tthinkthatyouarelookingatthisrationally. LikelyyourRAID6arraywouldhavediedheretoo. Thechanceforfailureonaparityarrayduringarebuildisverylarge. Thereissomuchtofail. flagReport Wasthisposthelpful? thumb_up thumb_down StorageNinja Thispersonisaverifiedprofessional. Verifyyouraccount toenableITpeerstoseethatyouareaprofessional. mace OliverSchwiegerwrote: Raid10Setupusesonly10drivesforRead/Write/IOPS. RAID10startswith4drives,andscalesinincrementsof2or4dependingonyourraidsub-system. Thereisnorequirementfor10drives... OliverSchwiegerwrote: Raid10Setupusesonly10drivesforRead/Write/IOPS.Theother10drivesarejustmirrorsoftheother10. False. Itcanreadfromall10drives,andwillhavethereadIOPSofall10drives.ItwillhavethewriteIOPSof5drives. OliverSchwiegerwrote: CPUsonthosearesofastthatparitycalculationsdoesntlowertheOverallPerformanceanymore. Itsnotaboutparitymath,itsaboutthefactthattheRAID5requires4WRITEoperationstothebankendspindles,whereRAID10requires2(Unlessyouhavesomesortoffancylogstructurelikestarwind/orVMwareVSAN,orothersystem"cheating"thewriteoverhead).Cachewillbluntthis,butitdoesn'tsolveitforsteadystatewrites. RAID10istwiceasfast. OliverSchwiegerwrote:= DellrecommendedRaid5/6SetupevenonthePowervault3220afterwehadseverePerformanceissueswithavirtualizedSQLServer. IfDellrecommendedRAID6foradatabase,unlessthatdatabasewas85%+readtheyscrewedup. DellstorageSE'sgenerallyknowbetterthesedays. OliverSchwiegerwrote: Intheendwemovedthedatabasetoaraid5on7SSDs+1ashotsparetogettheIOPSandreadPerformanceweneeded. A4:1writeoverheadisn'tabigdealwhenyouhave120KIOPSacrossthe8spindles. Spice(2) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin SWEManswrote: Ireadthisrecently,butIdon'tunderstandtheformula.AccordingtothisarticleRAID6willbenogoodin2019. http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805 Correct,RAID6doesnotscaleverywell. ParityRAIDjustdoesn't. Sowhenyourarraysgettoobig,astheysoonwillonaverage,RAID6willdoeffectivelynothingtoprotectthem. RAID10doesnothavethescalingissueandwillcontinuetoprotect. flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin NazihHaddadwrote: JustwonderingnowwhydoVMwareandsomeothersstillrecommendRaid5asachoice.isitjustforhavinglargerdiskspace? Sotwothings. OneisthattheydoNOTrecommendRAID5. Theyrequireitasaminimum. Thoseareverydifferentthings. Theyallowyoutobereckless. ThesecondthingisthatRAID5isNOTtheRAIDlevelyouendupwith,itisonlypartofit. WithVSAyouwouldendupwithRAID51or55,forexample. ThereisanotherlevelofRAIDyouarenotseeing. SoevenwhenRAID5isused,itismitigatedinthatscenario. RAID5isallowedbecauseatoofcustomersjustdemandit. LookatSpiceworks,tonsofpeoplewon'ttalkaboutanythingelse-math,logicorcommonsensebedamned. ItwouldbefoolishforVmwaretothrowawaycustomersbecausetheywanttoprotectthem. VMwaredoesn'tcareifyoulosedata. Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down ScottAlanMiller purecapsaicin OliverSchwiegerwrote: well,theydidn'ttellustheexactreason,butyourexplanationseemsreasonable. inthemeantimewechangedtheconfigfromraid5toraid10(10active200GBSSDplus1hotspare)anddidn'tseeanychangeinperformance.Westillconsiderchangingitback,oreventrytochangetheconfigtoraid6. Youeasilycannotpushtheperformancetestanyfurtherorhavesaturatedsomeotherbottleneck. Spice(1) flagReport Wasthisposthelpful? thumb_up thumb_down Prev12Next lock Thistopichasbeenlockedbyanadministratorandisnolongeropenforcommenting. Tocontinuethisdiscussion,pleaseaskanewquestion. Readthesenext... TheSOCBriefingforMay13th-Weareback!SecurityGoodmorningandwelcometotoday'sbriefing.Iamstartingthisback,usuallywilltrytopostmostlyFridaysonthisasmuchasIcan.AsusuallythispostbringvariouspostsregardingPatches&UpdatesplusSecurityNewsandabonusSecurityTip!Hope... Snap!Cybersecurity&theboardroom,Cooper,Starlinksatinternet,&SgrA*SpiceworksOriginalsYourdailydoseoftechnews,inbrief. NotonlyisitFriday,butitisalsoFridaythe13th!Adaythathasinspiredalate19th-centurysecretsociety,anearly20th-centurynovel,ahorrorfilmfranchise,andtriskaidekaphobia,awordIhadto...DellServerEatingPowerSuppliesHardwareIhaveaDellPowerEdge750that'sbeenquietlyrunningouroldaccountingsoftwaretonetworkusersforatleast15years.Theserverbeganpoweringoffintermittently,soIpickeduparefurbishedpowersupplyasareplacement.Itranfor3daysandfai...IE11retiresin1monthonJune15,2022WindowsReceivedthisreminderemailfromMSthismorning.Iwon'tbesadtofinallyseethisgo,eventhoughIknowIhaveafewholdoutswhoinsistonclickingonthatEstill(mostlyoutofhabit)IseveryoneReadyforthebigday?-----------------------------...Spark!ProSeries-May13th,2022SpiceworksOriginals Tomy fellowleft-handers,HappyFridaythe13th.Whiletheright-handed majorityquakesinfearandwallowsinsuperstition,amaladyknownasfriggatriskaidekaphobia,wecelebratethedaythatpromiseshealth,wea...



請為這篇文章評分?