ConversationalAvailabilityfortheAlways-On
Enterprise
ByNickCavalancia
Copyright©2015
ConversationalAvailabilityfortheAlways-OnEnterprisePublishedbyConversationalGeekInc.
www.conversationalgeek.com
Allrightsreserved.Nopartofthisbookshallbereproduced,storedinaretrievalsystem,ortransmittedbyanymeans,electronic,mechanical,photocopying,recording,orotherwise,withoutwrittenpermissionfromthepublisher.Nopatentliabilityisassumedwithrespecttotheuseoftheinformationcontainedherein.Althougheveryprecautionhasbeentakeninthepreparationofthisbook,thepublisherandauthorassumenoresponsibilityforerrorsoromissions.Norisanyliabilityassumedfordamagesresultingfromtheuseoftheinformationcontainedherein.
TrademarksConversationalGeek,theConversationalGeeklogoandJ.theGeekaretrademarksofConversationalGeek.Alltermsmentionedinthisbookthatareknowntobetrademarksorservicemarkshavebeenappropriatelycapitalized.Wecannotattesttotheaccuracyofthisinformation.Useofaterminthisbookshouldnotberegardedasaffectingthevalidityofanytrademarkorservicemark.
WarningandDisclaimerEveryefforthasbeenmadetomakethisbookascompleteandasaccurateaspossible,butnowarrantyorfitnessisimplied.Theinformationprovidedisonan“asis”basis.Theauthorandthepublishershallhaveneitherliabilitynorresponsibilitytoanypersonorentitywithrespecttoanylossordamagesarisingfromtheinformationcontainedinthisbookorprogramsaccompanyingit.
AdditionalInformationForgeneralinformationonourotherproductsandservices,orhowtocreateacustomConversationalGeekbookforyourbusinessororganization,pleasevisitourwebsiteatConversationalGeek.com
PublisherAcknowledgments
Allofthefolksresponsibleforthecreationofthisguide:
Author: NickCavalancia
ProjectEditor: J.PeterBruzzese
CopyEditor: JohnRugh
ContentReviewers: ShannaG.Giarrano
NotefromtheAuthor
Weallknowthesimpletruthofupisgood,downisbad.But,applicationscrash,hardwarefails,anddatabecomescorrupt–it’soneofthereasonswehaveajobinthefirstplace!
So,isitevenpossibletogetoperationstoanalways-onstatewhereavailabilityisthenorm?
I’lltrytoprovidethebestanswerpossible,butasyoucanseefromthelackofthicknessofthisbook,it’snotmeanttobeanexhaustivereferenceonestablishingavailabilityintheenterprise.Itwaswrittenwiththeintentofwalkingthroughtheneedforavailabilityandofreviewingtheimprovementsintechnologyandstrategythatassistingettingtoavailability.Itthenaimstoshowyouthatit’sreally,reallypossibletodaytoactuallyachievebothavailabilityofmission-criticalapplications,andtobecomeanalways-onenterprise.
So,grabacomfortablechair,andlet’sdiveintowhatit’sgoingtotaketogettothealways-onenterprise.
-NickCavalancia
The“Conversational”Method
Wehavetwoobjectiveswhenwecreatea“Conversational”book:First,tomakesureit’swritteninaconversationaltonesoit’sfunandeasytoread.Second,tomakesureyou,thereader,canimmediatelytakewhatyoureadandincludeitinyourownconversations(personalorbusiness-focused)withconfidence.
Thesebooksaremeanttoincreaseyourunderstandingofthesubject.Terminology,conceptualideas,trendsinthemarket,andevenfringesubjectmatterarebroughttogethertoensureyoucanengageyourcustomer,team,co-worker,friendandeventheknow-it-allBestBuygeekonalevelplayingfield.
“GeekintheMirror”Boxes
Weinfusehumorintoourbooksthroughbothcartoonsandlightbanterfromtheauthor.Whenyouseeoneoftheseboxes,it’stheauthorsteppingoutsidethedialogtospeakdirectlytoyou.Itmightbeananecdote,itmightbeapersonalexperienceorgutreactionandanalysis,itmightjustbeasarcasticquip,butpleasedon’tskipthese“geekinthemirror”boxes.
Greetings.I’mtheauthor’sinnergeek.WithintheseboxesIcansharejustaboutanythingonthesubjectathand.Read’em!
Demandforthe“Always-On”Enterprise
Howmuchdowntimecanyourbusinessreallyhandle?Ofcourse,youhaveanimmediateanswer.ButgoaskyourCEO–dotheyhaveadifferentanswer?Perhapsevenonethat’sabitunreasonable?I’massumingyou’dcomeupwithawell-thought-outnumberofminutesorhours,basedonyourabilitytorecover,Whereas,yourCEOwilllikelyjusttellyou“none.”
Soundaboutright?
There’sareasonwhyyourCEOisthinkingthatway(andit’snotlikethey’recrazy,orhaveunrealisticdemands,right?Thatneverhappens…).Thereasonthey’rethinkingaboutzerodowntimeisbecausetheexpectationofcustomers,partners,andevenemployeesisthatservicesyouprovidearealwaysavailable.
Whileitusedtobeacceptableforevenenterprisecompaniestobe“openforbusiness”from9to5,today’scustomerdemandstheabilitytoaccessandinteractwithacompany24/7viatheirwebsite,telephone,email,andevenchat.Purchases,accountupdates,accesstodata,support,andmoreanytimeyouwantisnowthenorm.
Now,youmightthinkit’sjustlargecompanieslikeTarget,BankofAmerica,orIKEAthathavetobealwayson–butyou’dbewrong.Thinkaboutit–thesamepersonwhoisexperiencingthisgreatanytime,anywhereaccesstoinformationandservicesfromthelargercompaniesisn’tgoingtoexcuseyoujustbecauseyou’renotaslarge.They’djustwonderwhyyoucan’tprovidethesamelevelofserviceinthisdayandage…and,perhaps,moveonifyoudon’t.
So,everyorganizationneedstobeupandrunningconsistently–andthat’snoeasytask!
An“always-on”enterprise,regardlessofsize,hasinvestedinamodernizeddatacenter,wherehardwareandoperatingsystemhavebeenkeptuptodate(sonoWindowsServer2003foryou!),andnewtechnologieshavebeenembraced.
Veeamrecentlypublishedtheresultsoftheir2014DataCenterAvailabilityReport,whichrevealsonly25%oforganizationsalreadyconsidertheirdatacentertobe“modern”.Lookformorestatsfromthisreportthroughout!
What’sinaModernDataCenter?Giventhestatabove,it’slikelyyoueitherdon’thaveamoderndatacenter,areworkingtowardsone,oraren’tsureifyoudo.Thelistbelowgivesyouanideaofthemorecommoninitiativesinvolvedinmodernizingthedatacenter.
• Servervirtualization• Upgradestostorageandoperatingsystems• Dataprotectionanddisasterrecovery• Networkvirtualization• VirtualdesktopInfrastructure(VDI)• Privatecloud• InfrastructureasaService(IaaS)• SoftwareasaService(SaaS)• DisasterRecoveryasaService(DRaaS)
It’saprettytalllist,sodon’tthinkyouneedtoimplementeveryinitiativeshown.Thisismoreaguidelinetoletyouknowifyou’reorganizationismovinginthedirectionof“modern.”
Doesamoderndatacenterequatetobeingalwayson?Notnecessarily.InVeeam’sAvailabilityreport,aboutathirdoforganizationsdidnotseebeingalwaysonasakeybusinessdriverformodernizingtheirdatacenter.
Ifyourorganizationisoneofthosethatareeithernotquitethere,orhaven’tstarted,giventhesomewhatdauntinglistabove,thequestiondoesneedtoberaised“Doyouevenneedamoderndatacenter?”
Therearemorereasonsthanjustavailabilitythataredrivingthemovetoamoderndatacenter.Organizationsthathavemodernizedtheirdatacenterarefocusingeffortsonthese
newertechnologiestoloweroperationalcosts,improvesecurity,andprovidemoreefficientservices.But,intheend,itallgoesbacktocustomerandemployeeexpectationsaroundaccessingapplicationsanddatanecessaryforthemboth,andthereforetheorganization,tobeproductiveandsuccessful.
End-userdemandisapowerfulmotivatorformovingtoalwayson.InVeeam’sAvailabilityreport,63%oforganizationscited“morefrequent,real-timeinteractionsbetweencustomers,partners,suppliersandemployees”asakeydriverforminimizingdowntime.
Thatbeingsaid,ifyoustrivetoprovideanalways-oninfrastructure,themoderndatacenterissomewhatarequirement.Asyou’llseeinthisnextsection,theevolutionfromjustbackupandrestoretotrueavailabilityreliesheavilyontheexistence(andimplementation)ofnewtechnologies.
TheEvolutiontoAvailability
Availabilitytodaymeansorganizationscanrecoverfromthelossofdata,applications,systems,orlocationswithminimalorzerodowntime.Thisisarealitythatissomewhatnew.Whileit’spossibletodaytocomereallyclosetothisdefinition(thezerodowntimeisthestickingpoint),toachieveavailabilitywe’vehadtoevolveabitfrombackuptechnologiesandmethodsofyesteryeartothemoderncapabilitiesweenjoytoday.
IfyougofarenoughbackinITlikeIdo,youcanrememberatimewhensimplyhavingdaily,weekly,andmonthlybackupswereconsideredenoughtoprotectthebusiness.Backupsranatnightafterbusinesshours,andyouusedasimplegrandfather-father-sontaperotationmethodtoestablishalong-termabilitytorecover.
IfyougobackasfarinITasIdo,youcananswerthefollowingquestion:
Q:WhatwasthemaindifferencebetweenDOSv3.2andv3.3?
Whenyoustopandthinkaboutwhatkindofprotectionanightly,weekly,ormonthlybackupreallyaffordedyourorganization,yourealize,atbest,itinvolvedthelossofatleast
A:v3.3brokethe33MBpartitionbarrier!
aday’sworthofwork.Fast-forwardtotodayandthis,obviously,wouldn’tcutitforanAlways-onEnterprise.
Recoveringbacktoyesterday?Seriously?
So,what’skeptyoufromgettingto“always-on”?
Therehavebeentwosetsofconstraintsovertheyearsstoppingorganizationsfromachievingtrueavailability(that,thankfully,havebeenslowlyerodingaway):
Thefirstistheavailabletechnology.Travelinthe‘WaybackMachine’afewyearsandtheonlywaytogethighavailabilitywasusingaclusteredserver.Today,wehavesomanymoreoptionstogetdata,services,andapplicationsmovedto,andrunningon,othersystemsinalternatelocations.
Thesecondisthecostofimplementingtheavailabilitystrategy.Thestrategyyouusedependsheavilyonthetechavailabletoyou.And,asyou’llsee,ourthinkingaroundwhat’spossiblewithregardtoavailabilityhaschangedthroughtheyears.
Let’slookateachoftheseandseehowtheybenefittedtheevolutionfrombackuptoavailability.
TheevolutionofbackuptechnologyBackuptechnologyusedtolimitustoasimpleonceadaybackupbecauseoftheconstraintsinthephysicalworld.Tapesystemsonlygosofast,harddrivescanonlypulldatasoquickly,backupprocessesimpactapplicationperformanceanduserproductivity,andschedulingdowntimeduringproductionusuallywasn’tanoption.
Enterdirect-to-diskbackups,thecloudasabackuptarget,andevenhybrid-cloudbackup,alongwithmoreadvancedmethodsofdetectingchanges,deduplication,andcompression–andthetechnologybegantoempowerthebackingupoffarmoredatamuchmorequicklywhile,somehow,stillmaintainingthe
abilitytorecoverthesameamountthatusedtotakeanentirenighttobackup.Youbegintoseehowimprovementsintechnologystarttoovercomethelimitationofthephysicalworldofservers,data,andbackup.
Abackuptapedriveinthe90’shada2:1compressionratioandoftenutilizedfurthersoftware-basedcompressiontoshrinkthebackupsize.Today’schangeidentification,compressionanddeduplicationmethodscanresultinaslittleas.1%ofthetotaldatasizebeingusedtocompleteanincrementalbackup.
Thenaddinvirtualization–acompletegamechangerforavailability,andyoucouldnowbackupchangesmadetoanentiresystem,ratherthanafile,application,ordatabase.Snapshottechnology,alongwithreplicationcapabilitiestooktheconceptofneedingtobackupasubsetofdataataparticulartimeandgaveITtheabilitytohaveaccesstoabackupthatprotectedtheentiremachine,suppliedmorebackupsatfasterintervals,andfacilitatedthecopyingofthosebackupstoanotherserver.Thishadareal“WOW”factortoit!
Andastechnologyevolved,gettingusclosertoavailabilityateachturn,sodidbackupstrategies.
TheevolutionofthecostofaprotectionstrategyIpreviouslymentionedtheGrandfather-Father-Sontaperotation–thatwasabackupandrecoverystrategythat,atthetime,wascuttingedgeandhadwithstoodthetestoftime.Itscostwaspredictable(asyouonlyneededafixednumberoftapes).But,liketechnology,recoverystrategieshaveevolved,eachonesuccessivelytakingadvantageofthelatesttechnologyandfindingaloweravailabilityprice-point.
Thefollowinglist,inmore-or-lesschronologicalorder,demonstratesthechangesinthinkingfrombackuptoavailability,astechnologyhasimprovedandcosthasreduced.
• BackupandRestore–thiswasatrulyviablemethodwhenfile-levelbackupsactuallyrepresentedtheentiretyofanoperatingsystem.Thefocushereissimplyfilelevel.Eventodaywhenbackingupvirtualmachineimages,thefocusofbackupandrestoreistosimplyrestorethatimagefile.
• BackupandRecovery–thisisashiftinthegoalwhileutilizinglargelythesametechnologyasbackup&restore.It’snolongeraboutjustrestoringthedatainthebackup,butreallyconcerningitselfwithwhetheryoucangettheintendedapplicationorsystembackintooperation.
• BusinessContinuity/DisasterRecovery(BCDR)–whenthisconceptwasintroduced,wecameclosetoavailability’sintentwiththislineofthinking.Therewasashiftfromjustrecoveringasystemorapplicationtotruly“howdowegeteverythingoperationalagain.”Theproblem?Itwasseenascostly,requiringexpensiveredundanthardware,useofcoldorhotsiteslyinginwait,comparativelyslowdeliverymechanisms,suchasUPSandFedEx,gettingonaplaneyourselftoworkontheremotesysteminperson,oraccessingaremotesystemviaslowWANlinkstogetcopiesofdatatooff-sitelocations.
• UseofVirtualization–byabstractingtheneedforactualhardware,virtualizationcreatedtheabilitytoinstantlyallowredundancythroughimageandsnapshotbackupsandreplicationtooffsitelocations.
Virtualizationalsomeantfarfewerphysicalservers,therebyreducingthecostofmaintainingasecondarysite.ThisadvancedtheBCDRmovement,givingitanewmeanstoreplicateentireenvironmentsfarmorequicklyandcheaply.
• UseoftheCloud–thecloudasastoragemediumhasbroughtsimple,inexpensive,andyetsecureoffsitebackupstothemasses.However,usingitforspinningupcomputeisstillrelativelynew,anditbringsnewpossibilitiesforhavingon-demandcomputeasaresource,leveragingtheuseofvirtualizationandreplication.
• DisasterRecoveryasaService(DRaaS)–Combiningtheuseofvirtualizationandthetrustinthecloud,buildingonthesimpleconceptofahotsite,DRaaSprovides(usually)manualfailoverservicestoareplicated,hostedcopyofanythingfromasingleservertoanentirenetwork.Fortier1applications,eventhislevelofservicecanbejustified,despitethehighcost.
• RecoveryasaService(RaaS)–whileDRaaStendstofocusonthereplicationandthenfailoverofservicesintheeventofadisaster,youcanthinkofRaaSasManagedDRaaS.Thatis,insteadofsimplyprovidingafailoverenvironment,RaaSinvolvestestingofreplicatedbackupstoensureviability,continuousrecoveryofdataatverysmalltimewindows,andautomatedfailover,allowingyoutomakeverycomplexenvironmentsliveinaveryfasttimeframe.
Bylookingthroughthisevolution,youcanseesomecommonthemes:
1) It’slessaboutdata&moreaboutcontinuity–Bytheendoftheevolutiontoavailability,nooneisreallyconcernedaboutwhichfilesarebeingrecovered;onlythatthecriticalsystemsandapplicationsareavailable.
2) Ashifttowardprotectinglargerdatasets–Thiscreatestheneedforidentifyingsmallerincrementalchangestobeprotected,ratherthantheentiredatasetasawhole.
3) Smallerbackupandrecoverywindows–Thisisaby-productofthesmallerincrementalchanges,butisakeyenablerofavailability.
4) Movefromprotectingjuston-premisestosomedegreeofoff-premisethinking–Sure,youarestillprotectingwhat’son-prem,butuseofthecloud,IaaS,SaaS,etc.areallpartoftheavailabilityplan.
5) Animprovedcost-to-benefitratio–Whilethelastoption(RaaS)probablycomesacrossasthemostexpensive(inalistthatissupposedtobegettinglesscostlyaswego),considerwhatthecostwouldbetoachievethisusingsimplebackupandrestore…andthenaddonthecostthelackofavailabilityofbackupandrestorebrings.Eventhoughthecheckwritteneachmonthisfarmorethanthecostofafewbackuptapes,thisoptionguaranteesyouarerecoverableinaveryshorttimeframe.
Allofthisputsussquarelywherewearetoday–thinkingaboutcontinualrecoveryprocessesutilizingbothon-premandcloud-basedinfrastructurestogetherinthemostcost-effectivewaytoempowertrueavailability.
Whileyouarelikelyinvestinginsomeofthetechnologiesthatmakeupamoderndatacenter,youprobablystillfindyourbackupandrecoveryeffortssomewhereinthemiddleoftheevolutiontoavailability.
So,howdoyoumakethetransitionfrombackuptoavailability?
Theanswerliesinfirstdefiningthedifferencebetweenthem.
Backupvs.Availability
Everyone,atsomepoint,stillperformsbackups.Theymightgobyanothername,suchassnapshotsorreplication.Butattheircore,theyservethesamepurpose.
Regardlessofwhereyouareinthestrategyortechnologyspectrumspreviouslylisted,whenitcomesdowntoit,youstillfocus,atsomelevel,onbackups.
Itmayseemthatavailabilityismerelyanadvancedformofbackupandrecovery(probablybecauseyou’rethinkingaboutitintermsofbackups).Inreality,nothingcouldbefartherfromthetruth.
Yousee,backups,attheircore,defineadatasetthat’sbeingprotectedandcanberecovered,andthenhaveyouworkforwardtoseeifwhatyou’vebackedupwillactuallygetyou
operational.Forexample,backupsofentireserversandsystemstatesinWindowsmaystillrequireanadditionalrecoverystep,suchasrestoringandbringingdatabasesinExchangeorSQLServerintoaconsistentstate.
Soasinglebackupdoesn’tnecessarilyequatetoarecoveredapplication,server,orenvironment.Ontheflipside,availabilitydemandsthatyoubeginwithwhatneedstobekeptrunningandworkbackwardstoidentifybackuptechnologiesthatmeettheneedandthebackupandrecoverystrategythatsupportstheendgoal.
So,justhowfarapartisyourbackupstrategyfromavailability?
Thereissomedifferencethatexistsbetweentherequirementsyourbusinesshasaroundkeepingsystemsandapplicationsavailable,andthebackup/recoverycapabilitiesyoucurrentlycanprovide.YoulikelyhaveidentifiedapplicationsthathavespecificSLAs,andyoualsoknowhowquicklyyourbackupsolutioncanrecoversaidsystem.Thenyoumultiplythatsamescenariobythenumberofcriticalapplicationsyouneedtorecover.Morelikelythannot,youcanrecoversomeoftheminthetimeallowed,butnotnecessarilyallofthem.
Othersofyoumaynothavedefinedtheavailabilityrequirements,soyoudon’tknowquitehowlargethegapbetweenbackupandavailabilityreallyis.
82%oforganizationsidentifythemselvesashavingagapbetweenbackupsandavailabilityintheAvailabilityreport.Ifyou’renotsure,youprobablyfitintothatgroup.
Solet’sgiveavailabilitysomelevelofdefinition.
Becauseyou’realreadyfamiliarwithbackups,let’sborrowsomecommontermsfrombackupsyou’refamiliarwith–after
all,evenwithavailability,youwillstillworkbackwardsallthewaytoabackupofsomekindatsomepointintheprocess.
Availabilityitselfisanobjective–agoaltobeupandrunningasclosetoalwaysaspossible.Backups,too,haveanumberofobjectivesthatspecifyoperatingparametersforyourbackups.Twospecificbackupsobjectivescanbeusedtohelpdefineavailability:RecoveryTimeObjectiveandRecoveryPointObjective.
RecoveryTimeObjective(RTO)istheamountoftimeallowedforrecoverytotakeplace.Thisusuallyisdefinedonaper-applicationorper-serverbasis,aseachapplication,system,ordatasethasitsownlevelofcriticalitytotheorganization.
RecoveryPointObjective(RPO)isthepointintimetherecoveryshouldbringthebusinessbackto.Or,inotherwords,theamountofdata(intermsoftime)theorganizationiswillingtolose.
So,iftheRTOisanhourandtheRPOis4hoursforagivenapplication,TheRTOdictatesyouhaveanhourtorecoverthegivendataset,whereastheRPOdictatestheITorganizationneedstobeplanningforbackupsofthatdatasetevery4hours,ataminimum.
So,whatshouldthevaluesbefortrueavailability?
Whilethereisnoindustry-definedvalue,agoodruleofthumbistoconsiderbothanRTOandRPOofjust15minutes.Iknow,Iknow…soundslikeI’macrazyperson,right?Butwe’retalkingaboutapplicationssoimportanttotheorganizationtheyneedtobetrulyavailable.ToparaphraseMiracleMax,“There'sabigdifferencebetweenmostlyavailableandavailable.Mostlyavailableisslightlydead.”
NoITproeverwantstotelltheexecutiveteamtheircriticalappis“slightlydead.”Byusingthe15minuteRTO/RPOvalues,you’dbeupandrunningin15minutesonlyhavinglost15
minutesofdata.That’saprettytightwindow,butyouneedtoconsidereachapplicationinyourenvironment,itscriticalitytothebusiness,andhowlongoperationscanreallygowithoutit.
Availabilityhasbeenaroundforyears,butthecostofachievinga15-minuteRTO/RPOhasbeentoohightomakeitcost-effective.
Regardlessofthespecificvaluesyoucomeupwith,thegoalshouldbetohaveverysmallvaluestoensurebothoperationsareupveryquicklyandverylittledatahasbeenlost.And,remember,ashortRTO/RPOispossible(moreonthatlater).
So,howdoyouachieveavailability?Isitjustbybuyingsomenewtechnology?Comingupwithanewstrategy?Orbydoingsomethingdifferent,instead?
AchievingAvailabilityHopefullybynow,yourecognizetheimportanceofgettingyourorganizationtoahighdegreeofavailability.Thesheernumberofidentifiablemissioncriticalapplicationsandsystemswithinyourorganizationalonewoulddictatehowimportantavailabilitytrulyis.So,howmanyisthat?Whatpercentageofyourapplicationandsystemworkloadswouldyouconsidermissioncriticaltotheorganization?
Giventhegrowthintheneedforavailability,woulditsurpriseyouatallthatthenumberofmissioncriticalapplicationsandsystemsinyourorganizationswillonlygrow,increasingtheneedforavailability?Thetimetogetthisrightisnow.
AccordingtoVeeam’sAvailabilityreport,almosthalfofallworkloadsareconsideredtobe“mission-critical”today.Thannumberwillincreaseby15%overthenexttwoyears.
Thereallygoodnewshereisyou’renolongerconstrainedbytechnology.TheabilityexiststodaytomeetverytightRTOsandRPOsbysimplytakingadvantageoftechnologiesthatarepartofamoderndatacenterstrategy.Now,itmaycostyoumoretoimplement,butgiventheimportanceofagivenapplication,thatcostmaybejustified.
Sincethetechnologyexists,achievingavailabilityisalotmoreaboutanexecutionstrategyandchoosingtherighttechnologytogetthere.Hereare5stepstohelpyoumakeprogresstowardachievingavailability.
1) Understandthebusinessrequirements–WhileI’vediscussedusingbackupobjectives,don’tstartthere(thosewillcomeinamoment);beginwithdefiningwhatthebusinessneedstooperate.Listouttheapplicationsthatarebothinternallyandcustomerfacingandcanbethoughtofasmissioncritical.
2) EstablishtiersofCriticality–whileeveryapplicationandsysteminthecurrentlistwillbe“critical”,weallknowyoucanweaseloutdifferenttiersofimportancefrommanagementifpressedhardenough.Afterall,ifyouwereinasituationwhereyoucouldonlygetoneofthemissioncriticalsystemsup,I’msuretheexecutiveteamwouldbeabletotellyouwhichonetheywant(andthen,ofcourse,they’dfireyouforonlybeingabletorecoverone…butIdigress).Thetiersof
importancemayverywellbealistofapplicationsindescendingorderofpriority,oryoumaybeabletoputthemingroupsbasedonequalimportance.
3) Developper-applicationrecoveryobjectives–Here’swhereyouputsomeactualnumberstoquantifywhatitmeanstobebothcriticalandatagiventier.Askthequestion“Howlongcanwebewithoutthisapplication,andhowmuchdatacanweaffordtolose?”todevelopRTOsandRPOs.Anyanswer,includingtheseeminglyunreasonable“Noneandnone”isanacceptableanswer.Why?Becauseitgivesyouagoaltohit(thatiswhythey’recalledobjectives)aswellasastartingpointwithwhichtobeginbuildinganavailabilitystrategy.
4) Buildastrategyaroundavailabletechnology–So,let’ssayyourERPsystemneedstohitthat15-minuteRTO/RPO.Howareyougoingtodothat?Lookatwhat’savailable(andIdon’tmeanjustwhatyouhavetoday,butwhatisavailableintheindustry)tocreateamethodbywhichyoucanachievetheobjectives.Witha15-minuteRPO,you’remostlikelygoingtoneedtolookatimage-basedbackupsofblockchangesreplicatedtoamirroredvirtualserver.Ifyou’relookingtoprotectagainstnaturaldisasters,thatserver’slocationwouldneedtoberemote.The15-minuteRTOmeanswhateverredundantinfrastructureyouhaveinplacewillneedtobelyinginwait,readytobefailedovertoatamoment’snotice.It’sthiskindofthinkingthat’sjustthestartofplanningyourstrategy.
5) Test,test,test–Thisistheoften-overlooked,mostcriticalstep.Imean,c’mon–we’retalkingaboutyour
mostcriticalapplicationsherewitha15-minutemarginoferror!Youneedtobecontinuously(notperiodically…continuously)testingeachandeverybackupthatisrecoveredtothealternatesitetoensure–withoutadoubt–thatwhenneeded,theapplicationwillrecover.
Somewherebetweensteps4and5is,ofcourse,implementingthetechnology.Ipurposelyleftthatout,asit’smoreatacticalaspect,ratherthanastrategicone.
Youcan’tjustbuysomething,switchitonand…BAM–Availability!It’sgoingtotakecarefulassessmentandplanningtocraftastrategyandexecutionplanthatwillensurethehighestlevelsofavailabilitypossible.
BeforeIendthissection,Iwanttocoverafewconsiderationsthatareamixoffinalthoughtsandnotesthatdidn’treallyfitneatlyanywhereelseinthebook,inanefforttogiveyouthebestchanceofmakingavailabilityhappen.
Availability-enablingconsiderations
• TheCloudandthe3-2-1rule–Acommonruleofthumbforbackupsistohave3copieson2differentmediatypeswith1ofthecopiesstoredoffsite.Thecloudprovidesanextremelycost-effectivemeanstoveryquicklymoveabacked-updataset(whetheritbefiles,aVMimage,etc.)offsite.Remember,disasterstakemanyforms,includingthelossoflocation,inwhichcasehavinganoffsitecopywillnotonlybehelpful,butnecessary.
• Goingoffsitedoesn’tguaranteeavailability–Whilehavinganoffsiteoption(whetherpublicorprivatecloud)helpsfromaDRandbackupperspective,
rememberthatoptionisstillonlyasgoodasyourcloudprovider.Besureyouknowwhoyou’resigningupwith,andgetboththeirSLAhistoryandaremediationplanshouldsomethinghappen.
• Backupsstillhavearole–Remember,availabilityisagoal;backups–eveniftheychangetheirspotsabitmovingfromfile-basedtoVMimage-based–arestillastrategicpartofthetacticalexecutionoftheavailabilitystrategy.
• Evenlocalstorage-basedsnapshotscanhelp–Thesebackupshavelessimpactonproductionsystemsandcanbecreatedfrequently.Whiletheyhavenorealplaceinthe3-2-1rule(astheysitinthesameplaceyourVMsare),theystillcanhelpbringaboutavailability,dependingonthetypeoffailure.
MakingAvailabilityaRealityIfthedemandtobealwaysonislowforyourorganization,don’tkidyourself–it’scoming.Eventually,everyorganizationwillfindsomelevelofneedforavailability,andtreatingthismatterinaproactivefashionnowwillhelpensureyoursuccess.
Today,thereisanabundanceoftechnologyabletosupporttheavailabilityofmissioncriticalapplicationsandsystems.Byimplementingsomeorallofthemoderndatacenterinitiatives,you’llalsobeputtinginplacethepiecesnecessarytobuildaproperavailabilitystrategythattakesadvantageofbackupswhilenotmakingthemthefocus.
Byidentifyingthebusinessrequirementsaroundavailability,establishingrelevantrecoveryobjectives,andworkingbackwardstotakeadvantageofthosepartsofthemodern
datacenteryou’veimplemented,you’llbewellonyourwaytoachievingastateofavailability.
Don’ttakeitallonatonce;startwithsomething…butstartnow.