timesboot.wim
boot.wim 时间:2021-05-24 阅读:(
)
UEFIPXEBootPerformanceAnalysisFebruary2014LiRuth,BrianRichardsonIntelCorporationUEFIPXEBootPerformanceAnalysisiiExecutiveSummaryNetworkbootusingthePrebootExecutionEnvironment(PXE)iswidelysupportedbycurrentUnifiedExtensibleFirmwareInterface(UEFI)implementations.
Inlargenetworkdeployments,bootperformanceiscritical.
Thisdocumentanalyzesfirmwareandoperatingsystem(OS)elementsthatimpactUEFInetworkbootperformance,andofferssuggestionsforreducingboottimeandremovingnetworkbottlenecks.
Thisdocumentfocusesonanaudienceoffirmwareengineers,BIOSvendors,networkengineersandODM/OEMsystemengineers.
UEFIPXEBootPerformanceAnalysisiiiContents1UsingPXEwithUEFI.
12FactorsAffectingUEFIPXEPerformance53TuningNetworkBootPerformance74PerformanceDatafromSamplePlatform95Summary131UsingPXEwithUEFIPrebootExecutionEnvironment(PXE)definesamethodforbootingcomputersusinganetworkinterface,independentoflocalstoragedevicesorinstalledoperatingsystems(OSs).
OnplatformswithUEFIfirmware,PXEissupportedbyanetworkstackintheclientfirmware.
Thenetwork'sDHCPprovidesapathtoabootserverandnetworkbootstrapprogram(NBP),downloadsitintothecomputer'slocalmemoryusingTFTP,verifiestheimage,andexecutestheNBP.
InaWindowsDeploymentServices(WDS)environment,theNBPisprovidedbywdsmgfw.
efi.
InaLinuxenvironment,theNBPisprovidedbyUEFI-enabledbootloaderssuchasGRUB,GRUB2orELILO.
ThisdocumentfocusesonbootperformanceforthePXEClientinUEFIfirmware,andhownetworktopologyimpactsPXEbootperformance.
BecausetheDHCPproxyprocesscontributesminimaloverheadtothenetworkbootprocess,analysisfocusesontheOSbootimagesdeliveredbyTFTP,whichcanbeupto200MBofdata.
1.
1UEFINetworkingTheUEFISpecificationdescribesaninterfacebetweentheOSandplatformfirmware,anddefinesageneralpurposenetworkstack,includingMNP,IP,TCP,UDP,andDHCPProtocols.
Tosupportbootfromnetworkdevices,thefirmwaremakesuseofUniversalNetworkDriverInterfaces(UNDIs)andprotocolsdefinedintheUEFISpecification:EFI_NETWORK_INTERFACE_IDENTIFIER_PROTOCOL(NII)EFI_SIMPLE_NETWORK_PROTOCOL(SNP)EFI_PXE_BASE_CODE_PROTOCOL(PXE)ThePXEprotocolisusedfornetworkaccessandbootingfromanetworkimage.
UEFIPXEBootPerformanceAnalysis2Figure1:UEFINetworkStackLayerFigure1illustratestheUEFInetworklayersrelatedtoPXE.
Thelowerhardwarelevel(UNDI/NII)isanabstractedinterfaceprovidedbythenetworkhardwarevendor.
ForIntelplatforms,theUNDIlayerisprovidedbyaUEFIDeviceDriverorbyaPCIExpressOptionROM.
TheSNPlayerwrapsaroundtheUNDIinterfaceandprovidesapacketlevelinterfacetothenetworkadapter.
TheMNPlayerprovidesarawasynchronousnetworkpacketI/Oservice,allowingmultipledriversandapplicationstousethesystemnetworkinterfacesatthesametime.
Themiddlelayerofthestackprovidesgeneralnetworkservices(IP,UDP,DHCPandMTFTP).
TheEFI_PXE_BASE_CODE_PROTOCOLislayeredontopofaUEFINetworkstackimplementation.
ThisprotocolalsoConsumesEFIMTFTP4andEFIMTFTP6protocolstoprovidetheTFTPservicesConsumesEFIDHCP4andEFIDHCP6protocolstoprovideDHCPservicesLayersoverEFIUDP4andEFIUDP6protocolstoprovideUDPReadandWriteservicestothenetworkinterface.
1.
2UEFIPXEwithMicrosoft*WDS*MicrosoftWDS*enablesPXEcompliantbootenvironments,startingwithMicrosoftWindowsServer2003R2.
Figure2isanoverviewoftheMicrosoft*WDS*PXEbootprocess.
UEFIPXEBootPerformanceAnalysis3Figure2:UEFIPXEwithMicrosoft*WDS*AfterSystemPoweron,theUEFIplatformfirmwareinitiatesthePXEbootprocess.
ThisrequiresfirmwaretoenabletheUEFINetworkStackandpresentanetworkbootoptionintheBootManagerMenu(BMM)orBootDeviceSelection(BDS)phase.
TheUEFIPXEBaseCodeprotocolrequestsanetworkaddressviaDHCP.
Microsoft*WDS*assignsanetworkaddresstotheclient,allowingtheclienttolocatethebootserverandrequesttheNBP(wdsmgfw.
efi).
OncetheNBPisdownloaded,theclientexecuteswdsmgfw.
efiasthefirststageofthenetworkbootprocess.
Thewdsmgfw.
efiexecutabletriggerfiledownloadforbootloader(bootmgfw.
efi),andbootloaderdownloadsadditionalfiles,includingbootconfigurationdatastoreandaRAMDISKimage(boot.
wim).
Theboot.
wimfileisover200MBforastandardclientsystem,sodownloadingthisfileconsumesmostofthePXEbootprocess.
Afterthesefilesaredownloadedtotheclientsystem,theWDSloadercallsExitBootService()tostopUEFIBootServicesandtransitiontoRuntime.
TheOperatingSystem(OS)kernelstartsexecutionandtakesoverhardwaremanagementfromUEFI,includingnetworkstackoperations1.
3UEFIPXEwithLinuxUnlikeMicrosoft*Windows*,LinuxdistributionshavetheoptiontosupportdifferentbootloaderswithPXEsupport.
ThebootprocessissimilartotheMicrosoft*WDS*implementation.
BootloadersusedbyvariousLinuxdistributionsincludeELILO,GRUBorGRUB2.
ThenextsectiondescribesusingELILOforUEFIPXEbootwithLinux.
UEFIPXEBootPerformanceAnalysis41.
3.
1ELILOFigure3UEFIPXEwithLinuxPXEServiceFigure3showsanoverviewoftheELILOPXEbootprocessusingUEFI.
It'sverysimilartotheMicrosoft*WDS*procedure,butwiththeLinuxbootloader(elilo.
efi),theconfigurationfile(elilo.
conf),andtheLinuxkernelimageandinitialramdisktakingtheplaceofthedownloadedfiles.
Thekernelandramdisksizesarevariable,dependingonconfiguration.
ConfiguringSUSESLE11SP3withabootimagefortheIntelDesktopBoardDQ77MKUEFI2.
3.
1DevelopmentKituseda4MBkerneland34MBinitialramdisk.
UEFIPXEBootPerformanceAnalysis52FactorsAffectingUEFIPXEPerformancePerformanceforUEFIPXEdeploymentsismeasuredusingclientboottimes.
ThisdocumentfocusesontwoaspectsofPXEbootperformance:EliminatingnetworkinfrastructureproblemsTuningnetworkbootperformancePXEreliesheavilyonTFTPforfiledownloads,andlargefilescancausedelaysifthenetworkisnotconfiguredproperly.
TuningtheTFTPconfigurationmayhaveasignificantimpactonnetworkbootperformance.
Thispaperalsodescribesamethodologyformeasuringbootperformance,andhowtoconfigureanappropriatetestenvironment.
2.
1MethodsforMeasuringNetworkBootPerformanceMeasuringbootperformanceforUEFIPXEissimilartomeasuringbootperformancefromalocaldisk.
Therearetwoboottimemeasurementstoexamine:Time1:TimebetweensystemresetvectorandReadyToBootevent(platforminitializationandbootenvironmentsetup)Time2:TimebetweenReadyToBootandExitBootServiceevent(hand-offtoOSfromtheUEFIBootLoader)SystemscompliantwiththeAdvancedConfigurationandPowerInterface(ACPI)SpecificationRevision5.
0canusetheFirmwarePerformanceDataTable(FPDT)torecordthisdata.
ACPIFPDTdataisrefreshedoneverysystemreset,soitisimportanttocollectthisdatapriortosystemresetorshutdown.
NetworkdownloadperformanceimpactstheTime2measurement.
TheplatformfirmwareusestheUEFINetworkStackfordownloadingbootfiles,basedontheTFTPprotocol.
TheMicrosoft*WDS*processreliesmoreontheUEFINetworkStack.
BecauseitdownloadslargerfilesviaTFTP,itistheprimaryexampleforournetworkperformanceanalysis.
ThedefaultconfigurationforTFTPstreamdepthisdynamicallyadjusted,butusingBcdEdityoucanadjusttheTFTPstreamdepthforbetternetworkperformance.
Thelargestfiledownloadedisboot.
wim,sothedownloadtimeforthisfileisusedasanindicatorfornetworkperformance.
Networksnifferapplicationsprovidemoreaccuratetimemeasurements.
UEFIPXEBootPerformanceAnalysis62.
2NetworkConfigurationNotethatnetworkbootperformancedataisonlymeaningfulinaspecificnetwork.
.
configuration,onewiththefollowingvariables:Networkbandwidth:Today'sIntelclientplatformsincludewiredGigabitEthernetControllers(1000Mbps).
ThebestnetworkperformancerequirescablingandswitchinfrastructuresupportingGigabitEthernet(GbE).
Networkswitch:Theperformanceofthenetworkswitchitselfhasanimpactonbootperformance.
SwitchesusedforbootperformancetestingshouldberatedforGbEbandwidth.
Certificationbyaninteroperabilityagencyishighlyrecommended(example:UNH-IOLtestingforIPv6andGigabitEthernet).
Serverconfiguration:Eventhoughthisdocumentfocusesprimarilyonfirmwareperformance,thebootserverconfigurationcanseverelyimpactPXEbootperformance.
TheTFTPprotocolrequiresinteractionbetweentheclientandserver,andoftentheclientmustwaitforaserverresponse.
Theserver'stransferratecanbeimprovedthroughoptimizingTFTPparametersandsizinghardwarerequirementstoproperlyservenetworkclients.
PXEclientconfiguration:Theclient'ssystemhardwareconfigurationcanaffectbootperformance,buttheplatformBIOSalsohasasignificantimpact.
ClientsystemsshouldusethelatestUEFIUNDIdriverfortheEthernetadapterandplatformfirmwarewithanupdatedUEFINetworkStack.
Networkbackgroundtraffic:Naturally,acrowdednetworkenvironmentwilldecreasetheoverallnetworkbootperformance.
ExcludingtheimpactofextratrafficforoptimalPXEbootperformancerequiresacleannetworkenvironment.
UEFIPXEBootPerformanceAnalysis73TuningNetworkBootPerformanceThissectionprovidesguidancetoBIOSVendorsandOEMandODMsystemdesignerstooptimizePXEbootperformance.
BIOSCustomization:EliminateUnusedNetworkModulesTheUEFINetworkStackprovidedbyEDKIIusesamodulardesign.
RefertotheUEFISpecification(sections2.
6.
2&2.
6.
3)forinformationonprotocolsrequiredfordifferentscenarios.
DriverProfileforPXEBootviaIPv4:PXEbootoverIPv4requiresthefollowingmodulesforacompletenetworkstack:SNPMNPARPIP4IP4CONFIGUDP4DHCP4MTFTP4PXEThesemodulesareprovidedbyEDKIIMdeModulePkg.
ThisalsorequirestheproperUNDIdriverforthenetworkinghardware,whichisprovidedbytheadapter'sPCIeOptionROMorUEFIDriver.
IfyouareusinganIntelEthernetAdapter,thelatestdriverscanbefoundatintel.
com.
DriverProfileforPXEBootviaIPv4andIPv6:IfbothIPv4andIPv6arerequired,thefollowingIPv6networkstackmodulesarerequiredfromtheEDKIINetworkPkg:IP6UDP6DHCP6MTFTP6.
AlsorequiredisthePXEdriverfromNetworkPkgwithsupportforbothIPv4andIPv6(NetworkPkg/UefiPxeBcDxe),buttheolderPXEdriverinMdeModulePkgmustberemoved.
UEFIPXEBootPerformanceAnalysis83.
1BootPolicy:EliminateNetworkImpactonSystemBootPerformanceAwell-designedBootDeviceSelection(BDS)policycanalsoimprovePXEbootperformance:Implement"fastboot"tooptimizethebootpath.
DonotconnectunnecessaryI/OdevicesinBDS.
IftheclientbootsfirstfromPXE,putthenetworkdevice'sbootoptionfirstintheUEFIbootorder.
ThisavoidsBDSconnectionstootherbootoptions.
Iftherearemultiplenetworkdevicebootoptions,placethepreferreddevicefirsttoavoidconnectingandconfiguringothernetworkdevices.
NetworkdevicesmaypublishtwoUEFIbootentriesperEthernetport:oneforIPv4andoneforIPv6.
PutthecorrectentryatthetopoftheUEFIbootorder,basedonthenetworktopology.
3.
2CodeRevision:UseUpdatedUEFIDriversTheentireUEFIPXEstackcontainstheEthernetadapterUNDIdriverandfirmwareUEFINetworkStackimplementation.
Updatingthesecomponentscanaddressknownperformanceissuesfromolderrevisions:Downloadthelatestdevicedriversfromintel.
comwhenusingIntelEthernetAdapters.
TheEDKIIprojecthasanopensourceUEFINetworkStackimplementationusingcomponentsfromtheNetworkPkgandMdeModulePkgpackages.
InteloffersUEFIDevelopmentKitplatformsforreference,basedonIntelmotherboardsrunningthelatestUEFIplatformfirmware.
TheseplatformsareusedforUEFItestinganddevelopment,andcontainupdatedversionsoftheIntelUNDIdriverandopen-sourceUEFINetworkStack.
MoreinformationcanbefoundattheIntelUEFICommunityResourceCenter.
3.
3BIOSPorting:PerformanceofSystemServicesTheUEFINetworkStackreliesonseveralfirmwareservices,whichmayslowbootperformanceifportedincorrectly:GetTime:TheGetTime()runtimeserviceistriggeredtoaddatimestamptoeveryTFTPDATApacket.
Thiscallinteractswiththeruntimeclock(RTC),and,ifnotoptimized,maydrasticallyslowTFTPdownloadspeeds.
GetVariableandSetVariable:TheUEFISpecificationrequiresnetworkdriverstoprovideNVRAMvariablesforconfigurationinformation(seeSection24andSection26forreference).
ThiscantriggermultiplecallstotheGetVariable()andSetVariable()servicesduringPXEboot.
MostvariableserviceimplementationsuseSystemManagementMode(SMM)forsecuritypurposes,whichcanaddoverheadifnotproperlyoptimized.
VerifytheSMRAMcacheisenabled(write-backmode)foroptimalperformance.
UEFIPXEBootPerformanceAnalysis94PerformanceDatafromSamplePlatformThischapterdetailstheperformancedatagatheredfromasampleenvironmentconfiguredusingtheIntelDesktopBoardDQ77MKUEFI2.
3.
1DevelopmentKit.
Thisisanisolatedtestnetwork,usingoneUEFIPXEClientandoneMicrosoftWDS*serverconnectedviaa1GbEswitch.
PXEClient:IntelDesktopBoardDQ77MKUEFI2.
3.
1DevelopmentKit(BIOSrevisionSDV.
MK.
B3),IntelCorei5-3550Processor,16GBDDR3RAM.
Connectedtothenetworkviaanon-boardGbEadapterMicrosoft*WDSServer*:Microsoft*Windows*2012,IntelPentiumDProcessor920,3GBDDRRAM.
ConnectedtothenetworkviaanIntelPRO/1000PMGbEadapter.
NetworkSwitch:HPProcurveSwitch6108(GbE)4.
1TFTPPerformancewithMicrosoft*WDS*Table1showsnetworkdownloadperformancedatausingTFTPwithMicrosoft*WDS*.
Thetimetodownloadboot.
wimfileisrecorded,countedfromthepointtheclientsendsaTFTP_READ_REQUESTforthefiletotheclientsendingTFTPACK,whichconfirmsreceivingthelastblock(averagedacrossthreePXEbootsessions).
TheSDV.
MK.
B3BIOSwastestedwithbothIPv4andIPv6.
Theboot.
wimfileisusedbecauseitisthelargestTFTPpayloaddeliveredtotheUEFIPXEclient.
Inourtestcase,theboot.
wimfilesizewas201MB.
Table1:Timemeasurementsfordownloadingboot.
wimPXEBootTestIWdsTransportTftpClientWindowSizeIWdsTransportTftpClientBlockSizeAverageTime(sec)IPv4(default)Dynamicadjusted145623.
7053IPv6(default)Dynamicadjusted145629.
0567IPv4(optimized)5014569.
5382IPv6(optimized)5014489.
3385Inourtestcase,Windows2012WDSstartedwiththefollowingparametervalues:UEFIPXEBootPerformanceAnalysis10ramdisktftpblocksize=1456ramdisktftpwindowsize=4ramdisktftpvarwindow=YesTheWindowsSizeandBlockSizeparametersarepartoftheBootConfigurationData(BCD)storeontheMicrosoft*WDS*server.
ThetestwasrunusingdefaultvaluesandoptimizedvaluestodemonstratehowserverconfigurationimpactstheTFTPtransferrate.
WhenusingthedefaultsettingfordynamicconfigurationofWindowSize,thevaluestartedat4andfinallyincreasedto64.
Wecaptured~4KTFTPACKpacketsduringtheprocessofdownloadingboot.
wimfile.
ThechangeinWindowSizevaluewascontrolledbyinteractionbetweentheloaderandWDSserver.
NotethattheBlockSizevalueforIPv6issmallerthanforIPv4,toavoidextraIPpacketfragmentation.
PleaserefertoMSDNformoreinformationonTFTPparameters.
4.
2ImpactofGetTime()andSetVariable()TheperformancecostforGetTime()andSetVariable()servicecanbemeasuredbymakingmultipleinterfacecallstotheseservicesandcalculatingtheaveragenumber.
ThemeasurementismadefromtheUEFIShellusingasimpleUEFIapplication.
TheapplicationcallstheGetTime()andSetVariable()functions1,000,000times,recordsthestartandendtime,thencalculatesandreportstheaveragecostforeachfunction.
UsingtheIntelDesktopBoardDQ77MKUEFI2.
3.
1DevelopmentKit(BIOSrevisionSDV.
MK.
B3),thecostforGetTime()is0.
0367microsecondsandthecosttosetaNon-NVVariable(1024bytes)throughSetVariable()is0.
04microseconds).
UsingthesametestmethodontheIntelServerBoardS2600CP4UEFI2.
3.
1DevelopmentKit(BIOSrevisionSDV.
CP.
B4),thecostforGetTime()is0.
0365microsecondsandSetVariable()is0.
168microseconds.
4.
3UEFIPXEBootPerformanceTable2showstheoverallUEFIbootperformanceusingtheACPIFPDTmethod.
Table2:PXEbootperformancetimePerformanceDataPXEbootoverIPv4(seconds)Normalboot(seconds)ResetEnd0.
0210.
021OSLoaderLoadImageStart7.
0375.
321OSLoaderStartImageStart20.
2465.
512ExitBootServicesEntry40.
3638.
861ExitBootServicesExit40.
3658.
862Thistablereportstimeforfivecheckpointsinthebootprocess:UEFIPXEBootPerformanceAnalysis11ResetEndOSLoaderLoadImageStartOSLoaderStartImageStartExitBootServicesEntryExitBootServicesExitInourtestcasethePXEBootprocessrequiresasystemreset,whichdestroyscurrentACPIFPDTdata.
ThismakesitimpossibletotestusingaUEFIShellApplicationoranOStool.
ToretrievetheFPDTdatabeforeitislostonreset,wereplacedtheFPDTmoduleintheRELEASEimagewithaDEBUGversion,includingamoduletooutputFPDTdataviaserialconsole.
Thebootperformancedatawascollectedfromtheserialloginformation.
ThefollowingexampleshowshowtoaddtheFPDTDXEdrivertotheplatform'sDSCfile:MdeModulePkg/Universal/Acpi/FirmwarePerformanceDataTableDxe/FirmwarePerformanceDxe.
inf{DebugLib|MdePkg/Library/BaseDebugLibSerialPort/BaseDebugLibSerialPort.
inf}Thetimebetween'ResetEnd'and'OSLoaderLoadImageStart'coversplatforminitializationandbootenvironmentsetup(Time1,insection2.
1).
.
Thetimefrom'OSLoaderLoadImageStart'to'ExitBootServicesExit'(Time2,insection2.
1)coversexecutionoftheDHCP/PXEserviceanddownloadingmultipleNBPimages.
NotethelargetimedifferencesbetweenthePXEandnormalboot(localdrive)scenarios.
ForamoredetaileddescriptionoftheFPDTfieldsseetheACPISpecificationRevision5.
0a('FirmwareBasicBootPerformanceDataRecord').
UEFIPXEBootPerformanceAnalysis12UEFIPXEBootPerformanceAnalysis135SummaryThisdocumentfocusedonPXEBootperformanceandprovidesdesignguidanceonhowtoobtaingoodPXEBootperformanceforBIOSdevelopers.
Tofacilitatethistopic,wedescribedthegeneralprocessofPXEboot,anddescribedthemethodofmeasuringnetworkperformance.
Welistedpossiblecomponentswhichimpactstheperformancedataandprovidedabenchmarkperformancedatausingonereferenceplatformandthesuggestedmethod.
UEFIPXEBootPerformanceAnalysis14AcknowledgementSincerethankstoFuSiyuan,ZhangChaoandWuJiaxinforhelpsettinguptestenvironments,workingthroughtestissuesandcollectingdataforthisdocument.
UEFIPXEBootPerformanceAnalysis15ReferenceDocumentsDocumentLocationUEFI2.
3.
1Specificationhttp://www.
uefi.
org/specsMicrosoft*WindowsDeploymentServiceshttp://msdn.
microsoft.
com/en-us/library/windows/desktop/dd379586(v=vs.
85).
aspxSUSELinuxEnterpriseServer11SP2forUEFIClientsBestPractices(WhitePaper)http://www.
novell.
com/site/docrep/2012/12/SUSE_Linux_Enterprise_Server_11_SP2_for_UEFI_Clients_Best_PracticesAdvancedConfigurationandPowerInterface(ACPI)Specificationhttp://www.
acpi.
infoPrebootExecutionEnvironment(PXE)Specificationftp://download.
intel.
com/ial/wfm/pxespec.
pdf[RFC2131]DynamicHostConfigurationProtocol,IETF,1997http://www.
ietf.
org/rfc/rfc2131.
txt[RFC2132]DHCPOptionsandBOOTPVendorExtensions,IETF,1997http://www.
ietf.
org/rfc/rfc2132.
txt[RFC3315]DynamicHostConfigurationProtocolforIPv6(DHCPv6),July,2003http://www.
ietf.
org/rfc/rfc3315.
txtTrivialFileTransferProtocol–TFTPhttp://www.
ietf.
org/rfc/rfc1350.
txthttp://www.
ietf.
org/rfc/rfc2347.
txthttp://www.
ietf.
org/rfc/rfc2348.
txthttp://www.
ietf.
org/rfc/rfc2349.
txtUEFIPXEBootPerformanceAnalysis16LegalDisclaimerINFORMATIONINTHISDOCUMENTISPROVIDEDINCONNECTIONWITHINTELPRODUCTS.
NOLICENSE,EXPRESSORIMPLIED,BYESTOPPELOROTHERWISE,TOANYINTELLECTUALPROPERTYRIGHTSISGRANTEDBYTHISDOCUMENT.
EXCEPTASPROVIDEDININTEL'STERMSANDCONDITIONSOFSALEFORSUCHPRODUCTS,INTELASSUMESNOLIABILITYWHATSOEVER,ANDINTELDISCLAIMSANYEXPRESSORIMPLIEDWARRANTY,RELATINGTOSALEAND/ORUSEOFINTELPRODUCTSINCLUDINGLIABILITYORWARRANTIESRELATINGTOFITNESSFORAPARTICULARPURPOSE,MERCHANTABILITY,ORINFRINGEMENTOFANYPATENT,COPYRIGHTOROTHERINTELLECTUALPROPERTYRIGHT.
Intelproductsarenotintendedforuseinmedical,lifesaving,orlifesustainingapplications.
Intelmaymakechangestospecificationsandproductdescriptionsatanytime,withoutnotice.
Thispaperisforinformationalpurposesonly.
THISDOCUMENTISPROVIDED"ASIS"WITHNOWARRANTIESWHATSOEVER,INCLUDINGANYWARRANTYOFMERCHANTABILITY,NONINFRINGEMENT,FITNESSFORANYPARTICULARPURPOSE,ORANYWARRANTYOTHERWISEARISINGOUTOFANYPROPOSAL,SPECIFICATIONORSAMPLE.
Inteldisclaimsallliability,includingliabilityforinfringementofanyproprietaryrights,relatingtouseofinformationinthisspecification.
Nolicense,expressorimplied,byestoppelorotherwise,toanyintellectualpropertyrightsisgrantedherein.
IntelandtheIntellogoaretrademarksorregisteredtrademarksofIntelCorporationoritssubsidiariesintheUnitedStatesandothercountries.
*Othernamesandbrandsmaybeclaimedasthepropertyofothers.
Copyright2014IntelCorporation.
Allrightsreserved.
gigsgigsCloud日本东京软银VPS的大带宽配置有100Mbps、150Mbps和200Mbps三种,三网都走软银直连,售价最低9.8美元/月、年付98美元。gigsgigscloud带宽较大延迟低,联通用户的好选择!Gigsgigscloud 日本软银(BBTEC, SoftBank)线路,在速度/延迟/价格方面,是目前联通用户海外VPS的最佳选择,与美国VPS想比,日本软银VPS延迟更...
gcorelabs提供美国阿什本数据中心的GPU服务器(显卡服务器),默认给8路RTX2080Ti,服务器网卡支持2*10Gbps(ANX),CPU为双路Silver-4214(24核48线程),256G内存,1Gbps独享带宽仅需150欧元、10bps带宽仅需600欧元,不限流量随便跑吧。 官方网站 :https://gcorelabs.com/hosting/dedicated/gpu/ ...
今天中午的时候看到群里网友在讨论新版本的Apache HTTP Server 2.4.51发布且建议更新升级,如果有服务器在使用较早版本的话可能需要升级安全,这次的版本中涉及到安全漏洞的问题。Apache HTTP 中2.4.50的修复补丁CVE-2021-41773 修复不完整,导致新的漏洞CVE-2021-42013。攻击者可以使用由类似别名的指令配置将URL映射到目录外的文件的遍历攻击。这里...
boot.wim为你推荐
政协晋城市委员会主办重庆网通中国联通重庆分公司的公司简介css3圆角如何用CSS实现圆角矩形?netbios端口26917 8000 4001 netbios-ns 端口 是干什么的ipad如何上网iPad怎么上网?请高手指点google图片搜索如何用google搜索空间照片google分析如何添加google analysis重庆电信测速重庆电信对BT开始限制了?morphvoxpro怎么用morphvox pro和全民k歌怎么用chrome18怎么关闭chrome的自动更新,稳定版要18了,mactype要悲剧了
郑州虚拟主机 jsp虚拟主机 武汉域名注册 中文国际域名 cybermonday 日本软银 a2hosting 外国服务器 紫田 网通ip 合肥鹏博士 日本bb瘦 微软服务器操作系统 如何注册阿里云邮箱 香港亚马逊 万网主机 宿迁服务器 江苏双线 rewritecond 新疆服务器 更多