CPUacceptencoding

F5Networks,Inc.
-1-Aug-05TheBIG-IPSystemWithIntelligentCompression:CuttingApplicationDeliveryTimeandOptimizingBandwidthOverviewThenumberandcomplexityofapplicationsdeliveredovertheInternetcontinuestogrow.
AvailablebandwidthisstretchedtocapacitybythesheervolumeandchattinessofInternetprotocols.
Theresultsarelongdelays,increasedlatency,unsatisfactoryend-userexperiences,andunacceptableapplicationperformancethatcostsorganizationsrealdollarsintermsofcustomerreputationandproductivitylosses.
Compressiontechnologycanprovidedramaticapplicationperformanceimprovements.
Thiswhitepaperdiscussestheneedforcompression,differentapproachesforcompressingwebtrafficavailableinthemarkettoday,andhowF5'sBIG-IPsystemprovidesorganizationswithapowerfulwaytooptimizetheirbandwidthintelligentlywhileacceleratingthedeliveryoftheirapplications.
WhatisCompressionCompressionisanoptimizationtechniqueusedtoremoveredundantpatternsfromadatastreamsothatithasfewerpacketsandconsumeslessbandwidth,significantlyimprovingapplicationperformance.
Sincetherearefewerpacketstraversingthenetworkfromtheservertotheenduser,applicationdatagetsdeliveredfaster.
HTTPcompressioniscommonlyusedforwebapplicationswhichhelpsreducethebandwidthconsumedbywebobjectsandsignificantlyimprovesenduserresponsetimes.
GZIPisapopularHTTPcompressiontechniqueappliedtowebtrafficandissupportedbystandardbrowsers.
MostbrowsershavebeenequippedtosupporttheHTTP1.
1standardknownas"content-encoding.
"WithGZIP,theclientbrowseressentiallynegotiateswiththeserverindicatingthatitcanacceptencodeddata.
Uponsuccessfulnegotiation,GZIPcompressesdatabeingsentoutfromthewebserverusingtheencodingformatacceptedbytheclient.
Inthefiguresbelow,Message1showstheclientrequestforcompressionandMessage2showstheserveracceptingtheGZIPcompressionrequest:GET/HTTP/1.
1Host:www.
f5.
comUser-Agent:Mozilla/5.
0(Windows;U;WindowsNT5.
1;en-US;rv:1.
5)Gecko/20031007Firebird/0.
7Accept:text/xml,application/xml,application/xhtml+xml,text/html;q=0.
9,text/plain;q=0.
8,video/x-mng,image/png,image/jpeg,image/gif;q=0.
2,*/*;q=0.
1Accept-Language:en-us,en;q=0.
5Accept-Encoding:gzip,deflateAccept-Charset:ISO-8859-1,utf-8;q=0.
7,*;q=0.
7Keep-Alive:300Connection:keep-aliveMessage1:ClientRequestforCompressionHTTP/1.
1200OKDate:Thu,04Dec200316:15:12GMTServer:Apache/2.
0Vary:Accept-EncodingContent-Encoding:gzipCache-Control:max-age=300Expires:Thu,04Dec200316:20:12GMTX-Guru:basic-knowledge=0,general-knowledge=0.
2,complete-omnipotence=0.
99Content-Length:1533Content-Type:text/html;charset=ISO-8859-1Message2:ServerResponseAcknowledgingCompressionUseF5Networks,Inc.
-2-Aug-05Oncethecompressiontypeisestablished,thesessionwillbeginconversingusingtheagreeduponcompressiontechnique.
Thebrowserwilldecompressdatareceivedbytheserveronthefly,reducingtheamountofdatasentandincreasingthepagedisplayspeed.
Therearetwowaystocompressdatacomingfromawebserver--dynamicallyandpre-compressed.
DynamicContentAccelerationtypicallycompressestransmissiondatasuchasHTML,XML,CSS,Java,JavaScript,andWMLonthefly.
DynamicContentAccelerationisparticularlyusefuline-commerceapplications,database-drivensites,andmanyotherapplications.
Pre-compresseddataistextbaseddatathatisgeneratedbeforehandandstoredontheserverinhtml.
gzfilesorotherformats.
Sincethedataispre-compressed,CPUloadisreducedascompression/decompressionisnotdoneonthefly.
ChallengeWhyisCompressionNeededThevolumeofwebtraffictriplesalmosteveryyearasapplicationsbecomeincreasingly'webified'.
Withthisexponentialincreaseindemandforbandwidth,combinedwithlowbandwidthclientconnectiontypes,organizationsareexperiencinghighnetworklatencyandpoorapplicationresponsetypes.
Tosolveapplicationperformanceproblems,thetraditionalsolutionfornetworkmanagershasbeentothrowmorebandwidthattheproblem.
Thisapproachdoesnotscalewellasitrequiresnetworkmanagerstorepeattheprocessasapplicationneedsgroworwhenthenext'big'applicationgetsdeployed.
Additionally,bandwidthpriceshavenotdeclinedasrapidlyasexpectedandaddingnetworkcapacityismoreexpensivethanimplementingcompression.
Let'slookinmoredetailatsomeofthedriversforcompression:Clientaccessspeedsandthelastmile:Duetothedifferentconnectivityaccessmethods,varyingbandwidthavailabilityandcomplexityoftheroutingprotocols,applicationtrafficissubjecttomanyconstraintswhichcauseitsperformancetodegradeovertime.
Thecascadingandmultipleeffectsofallthesevariablesmakestheapplicationof"smartcompression"extremelyuseful.
Forinstance,becauseoflow-bandwidthavailabilityandhighlatency,dial-upor'satelliteclients'experiencetheworstresponsetimeswhichcanliterallyrenderanapplicationunusable.
Theabilitytoonlycompressdial-upandsatellitetrafficwilldramaticallyreduceCPUloadingrequirementsontheswitch,allowingforgreaterefficiencyatalowerprice.
Organizationsthatinadvertentlycompresstrafficfrombroadbandclientsmaypreventoptimizingthebenefitsofcompression.
Networkthroughputandbandwidthlimitation:Organizationsarefacedwiththechallengeofusingtheirexistingbandwidthintelligently.
Webapplicationobjectsizeshavebeensteadilygrowingandorganizationsfindthemselvesencounteringbandwidthbottlenecksonaregularbasis.
Thishastheeffectofslowingdownotherapplicationsastheygetstarvedforbandwidthwhichintroducesveryhighlatency.
Delay-sensitiveapplicationslikeVoIPcannottoleratesuchhighlatencyandbecomeunusable.
WhereDoICompressMyApplicationTrafficAlthoughthebenefitsofcompressionareobvious,thelocationwherecompressionisappliedtoapplicationtrafficplaysaveryimportantroleindeterminingtheefficacyofthecompressiontechniqueandtheoverallimprovementinapplicationdeliverytime.
Therearetwomainstreamapproachestocompressingwebapplicationtraffic:CompressionontheserverThiscommonlyusedapproachinvolvescompressingallapplicationtrafficbeforeitleavestheserver.
Theservercanbeusedtocompressstaticaswellasdynamiccontent.
ManypopularwebserverssuchasMicrosoft'sIISandApachesupportserversidecompression.
Thisapproachtocompressionworkswellforasmallnumberofapplicationsbutdoesnotscalewellasthenumberofapplicationsandthesizeofapplicationobjectsgrow.
Also,thisapproachmakesithardertochangeoraddcompressionparametersasthishastobedoneacrossmultipleservers,increasingmanagementoverhead.
CompressionisaveryprocessorF5Networks,Inc.
-3-Aug-05intensivefunction.
CompressingapplicationtrafficontheserverconsumesvaluableCPUcyclesanddegradesserverperformancebyintroducinglatencyinapplicationresponsetimes.
Toexacerbatetheproblemfurther,morecomponentsofapplicationdeliverysuchasSSLencryption,clientauthentication,security,etc.
arealsobeingrunontheservers,addingevenmoreoverheadontheCPU.
Astheserversbecomeachokepoint,thebenefitsofcompressionarenotfullyrealizedandapplicationdeliverytimeswillfailtoshowadramaticimprovement.
Figure1:CompressionontheserverCompressionontrafficmanagementdevicesAnewmeanstocompressioninvolvesoffloadingcompressionontoanapplicationtrafficmanagementdevicethatfront-endstheservers.
Applicationtrafficcannowbecompressedbythedevicesonbehalfofthewebserver,thusremovingtheCPUbottleneckimposedbycompressiononthoseservers.
Thisapproachisgaininginpopularitysinceorganizationscannowrealizethefullbenefitsofcompression.
Therearetwowaystoachievethis:oSymmetric:Thisapproachinvolvesoffloadingthecompressionfunctiontoadevicedeployedinfrontofthewebserverswithaseconddeviceorasoftwarecomponentattheclientend.
Thisend-to-endproprietaryapproachissuitedforabranchofficedeploymentbutdoesnotworkwellforwebapplicationsthatareaccessedbymillionsofclientsontheInternet.
Deploymentandadministrationofthistypeofsolutioniscumbersome,asdownloadingorpushingapieceofsoftwarefromtheservertotheclientisintrusiveandmaycompromisesecurity.
oAsymmetric:Thisapproachinvolvesoffloadingthecompressionfunctiontoatrafficmanagementdevicethatcanbedeployedinfrontofwebservers.
Thedecompressionisdonebybrowsersattheclientend.
Thisapproachtakesadvantageoftheexistingdecompressioncapabilitiesthatareapartofallstandardbrowsersandrequiresnochangesattheclientend.
Italsoeliminatesthemajorityofbrowsercompatibilityissuessincethedevicenowactsasamediatorandtranslatesbetweentheclientbrowserandserver.
ThisapproachisgainingpopularityasorganizationsareabletorealizethefullbenefitsofcompressionbysavingvaluableCPUcyclesontheirserversaswellasobviatingtheneedforanychangestotheclientinfrastructureorintrusivedownloads.
Thisapproachalsoallowsorganizationstocentralizetheirmanagementofallcompression-relatedconfigurationsandeliminatesoverheadthatisassociatedwithmanagingmultipleservers.
F5Networks,Inc.
-4-Aug-05Figure2:CompressionontheBIG-IPdeviceSolutionF5'sBIG-IPCutsApplicationDeliveryTimeandOptimizesBandwidthF5'sBIG-IPapplicationtrafficmanagementsystemprovidestheindustry'smostscaleable,intelligentandflexiblecompressionsolution.
ByasymmetricallyoffloadingHTTPcompressionfromservers,theBIG-IPsystemreducesserveroverheadanddecreasestotalcostofownershipforserversupto65%byenablingserverconsolidation.
TheBIG-IPsystemtakesadvantageofexistingbrowserdecompressioncapabilitiesandobviatestheneedforanyclientsidechangesorintrusivedownloads.
BIG-IP'sIntelligentCompressionusesapatentpendingapproachtodetermineclientconnectionlatency,decreasingbandwidthusageby60-80%whileimprovingenduserresponsetimesbyover200%.
TheBIG-IPsolutionisthefirstintheindustrytoprovideorganizationswithascaleablecompressionsolutionwiththeoptionofcompressingwebtrafficthroughoptimizedhardwarewithitsAdaptiveCompressionOffloadfeature.
WhatIsIntelligentCompressionTheBIG-IPsystem'sIntelligentCompressionprovidesorganizationsawaytotargetusersforcompression.
Compressingalltypesoftrafficdoesnotnecessarilyyieldareductioninbandwidthusage.
Thechallengewithcompressionistoknowhowtotargetitmostefficiently,sousersreceivethemostbenefit.
Forexample,adial-upclientwillbenefitmostfromthecompressionsinceithashigherlatencywhilethebenefittoabroadbandclientwouldbeinsignificant.
Thisisbecausebroadbandclientshavelargerreceive-windowsizes.
Compressioncausesresponsetimestodegradeasclientsnowhavetowaitlongertoreceivedata,negatingsomeofthebenefitsofcompression.
BIG-IPusesapatentpendingtechniquetodynamicallydeterminetheclientconnectionlatency.
TheBIG-IPsystemmonitorstheTCPRTT(RoundTripTimes)todynamicallycalculateuserlatency,allowingBIG-IPtodevotemorepowertocompressingtraffictouserswhoneeditmost.
TunableCompressionEngineBIG-IPprovidesorganizationsthecapabilitytofinetunetheircompressionparameterstogetthemaximumbenefit.
Theycantargetsystemresourcesforcompressionwheretheymattermost(achievingahigherROI)aswellasachievingbettercontrolwhencompressionisinvoked.
Thefollowingcompressionparameterscanbeconfigured:Minimumcontent-lengthforcompressionThisparameterspecifiestheminimumlengthoftheserverresponse(inbytes)tobeconsideredacceptableforcompression.
TunablecompressionThisparameterallowsspecificationofthedesiredcompressionratio,CPUandmemoryutilization.
F5Networks,Inc.
-5-Aug-05F5'sAdaptiveCompressionOffloadenablesorganizationsto"tune"compressionwhichboostsbandwidthsavingsbyfreeingupvaluablesystemresources.
ByadaptivelyoffloadingHTTPcompressionfromthesystemprocessortotheoptimizedhardwarewhenthesystemCPUreachesacertainthreshold,BIG-IPcandeliverunprecedentedlevelsofcompressionthroughput(upto2Gbps)andminimizesystemprocessoroverheadupto80%.
GranularL7PolicyBasedCompressionTheBIG-IPsystemprovidestheindustry'smostgranularsolutiontocontrolhowandwhattypeoftrafficiscompressed,deliveringbetterperformanceandimprovedROI.
Organizationscanconfigurecompressionpervirtualserver,sourceIP,destinationIP,filetypeorprotocol,orbasedonanyLayer7variable.
ByleveragingtheBIG-IPsystem'suniqueiRulescapability,customerscanchoosetoenablecompressionbasedonthesegranularL7inspectioncriteria,allowingorganizationstoturncompressiononoroffforanindividualHTTPrequest/responsepair.
ContentFiltersandExceptionHandlingBIG-IPalsoprovidesorganizationspredefinedfiltersthattheycanusetotargetdesiredcontenttypesandalsohandleexceptions.
Inorderforaserverresponsetobecompressed,usersmaydefine"include"and"exclude"liststobettertargetcompressionorquicklyhandleexceptions.
Suchpredefinedfilteringcapabilityincludes:URI(fromtheclientrequest)ThisisalistofregularexpressionsusedtomatchtheRequest-URIpartoftheclientrequestline.
Forexample,toincluderequestsendingin".
txt",".
htm"and".
html",onewouldusethefollowingintheURIincludefield:".
*\.
txt"".
*\.
htm"".
*\.
html".
ContentTypes(fromtheserverresponse)ThisisalistofregularexpressionsrepresentingMIMEtypes.
Theregularexpressionswillbecheckedagainstthevalueoftheserver's"Content-Type:"header.
Forexample,todisablecompressingPDFsandallimagefiles,onewouldusethefollowinginthecontenttypeexcludefield:"application/pdf""image/.
*".
Toincludealltexttypes,onewoulduse"text/.
*"inthecontenttypeincludefield.
Toincludeallnon-CSStexttypes,youwoulduse"text/(!
css$)"inthecontenttypeincludefield.
VisibilityIntoCompressionPerformanceTheBIG-IPdeviceprovidesrichstatisticstomonitorcompressionperformanceandtodemonstratekeybenefitstotheorganization.
Thesestatisticshelporganizationstunetheircompressionpoliciesandmeasurethebenefitsofcompressiondependingonthetypeofobjectbeingcompressedandprovidevisibilityintobandwidthsavings.
Thesestatisticsinclude:SizeoftheobjectbeforecompressionSizeoftheobjectaftercompressionObject-typebeingcompressedF5Networks,Inc.
-6-Aug-05Figure3:BIG-IPcompressionstatisticsROIExampleCompressionROIhasadirectimpactoncostthatisrelativelyeasytocalculate.
Letusconsideranorganizationthathasthefollowingbandwidthpricingstructure:Figure4:ExamplebandwidthpricingstructureAssumingtheorganizationexperiencestrafficspikesandbandwidthusageof2.
5MbpsthatgoesoverintoTier2,themonthlychargesare$3500.
Iftheorganizationistobeabletoreducethebandwidthusagebelow2Mbps,however,theorganizationwon'tgetpenalizedeverymonthforbandwidthoversubscription.
Compressioncanhelpthecompanyreachthisobjective.
BIG-IPprovidesvariouscompressionmodulesatdifferentbandwidthrates(5Mbps,100Mbps,500Mbps,1000Mbps)tomatchthecustomerneeds.
Inthecaseabove,deploymentofaBIG-IPcompressionsolutionresultedinthefollowingROI:Figure5:ROIresultsusingcompressionontheBIG-IPdeviceF5Networks,Inc.
-7-Aug-05ConclusionTheBIG-IPdevice'sIntelligentCompressioncapabilityprovidesamarketleadingapproachtooptimizebandwidthandacceleratethedeliveryofapplicationsfromtheservertotheend-user.
Withincreasingapplicationperformancechallenges(bandwidthbottlenecks,delays,timeoutsandoutages),lossofrevenueandcustomerdissatisfactionisbecomingcommonplaceandtheneedtosolvetheseproblemshasbecomeimperative.
BIG-IP'scompressionsolutionenablesorganizationstosolvetheirapplicationperformanceproblemsbydeliveringthefollowingbenefits:TheBIG-IPsystemusessophisticatedbandwidthoptimizationtechniquessuchasIntelligentCompressiontoreducelatencyandimproveend-useraccessandpagedownloadtimesbyover200%andimproveperformancebydecreasingbandwidthusageby60-80%.
Unlikethelegacyserver-sidecompressionmodel,BIG-IPoffloadsserveroverheadanddecreasestotalcostofownershipofserversupto65%.
Unlikethelegacyserver-sidecompressionmodel,BIG-IPcentralizescompressionmanagementandeliminatesbrowserincompatibilitiesbyactingasamediatorbetweentheclientandtheserver.
Unlikesymmetriccompressiondevices,BIG-IP'sHTTPcompressioninherentlytakesadvantageofthedecompressioncapabilitiesontheclientbrowserandobviatestheneedforanychangestotheinfrastructureontheclientside.
BIG-IPprovideoptionalcompressionoffloadASICswhichworktofurtherscaleandoffloadcompressioncyclesfromtheinfrastructure.
UsedinconjunctionwithBIG-IP'smultiplexing,cachingandTCPoptimizationfeatures,BIG-IP'scompressiondeliversaddedenduserperformanceimprovementandbandwidthavailability.
AboutF5F5enablesorganizationstosuccessfullydeliverbusiness-criticalapplicationsandgivesthemthegreatestlevelofagilitytostayaheadofgrowingbusinessdemands.
AsthepioneerandgloballeaderinApplicationTrafficManagement,F5continuestoleadtheindustrybydrivingmoreintelligenceintothenetworktodeliveradvancedapplicationagility.
F5productsensurethesecureandoptimizeddeliveryofapplicationstoanyuser-anywhere.
Throughitsflexibleandcohesivearchitecture,F5deliversunmatchedvaluebydramaticallyimprovingthewayorganizationsservetheiremployees,customersandconstituents,whileloweringoperationalcosts.
Over9,000organizationsandserviceprovidersworldwidetrustF5tokeeptheirbusinessesrunning.
ThecompanyisheadquarteredinSeattle,Washingtonwithofficesworldwide.
Formoreinformationgotowww.
f5.
com.