揭开超大规模队列管理的带内和带外方法的神秘面纱.pdf

1、Demystifying In-band/Out-of-band Methods for Hyperscale Fleet ManagementFuture Technology InitiativesDemystifying In-band/Out-of-band Methods for Hyperscale Fleet ManagementGeorge Kola,Senior Staff SW Engineer,GoogleAnil Agrawal,HW Systems Engineer,MetaAgendaHyperscale Fleet Deployment ModelsHardwar

2、e Management ScopeIn-band/Out-of-band Management MethodsQ&AHyperscale Fleet Deployment ModelsHyper Scaler ownership varies based on the Deployment ModelHardware Management Methods Varies(In-band and Out-of-band)CustomerHyper ScalerEnd UserOperating System(OS)FirmwareHardwareServicesApplications(SW)B

3、are-MetalHypervisor(optional)CustomerHyper ScalerEnd UserOperating System(OS)FirmwareHardwareServicesApplications(SW)InfrastructureHypervisor(optional)Hyper ScalerEnd UserOperating System(OS)FirmwareHardwareServicesApplications(SW)ApplicationHypervisor(optional)System Configuration(E.g.,FW Upgrades/

4、downgrades,etc)Hardware ManagementHardware Management ScopeSystem Monitoring(E.g.,FW versions,PCIe device inventory,sensors,etc)Hardware Fault Management(e.g.,Core,Cache,memory,PCIe,internal error handling,etc)Keeping the Scope of Discussion System Monitoring and Hardware Fault ManagementIn-band and

5、 Out-of-band ManagementIn-bandIn-bandOut-of-bandOut-of-bandHyper Scaler Management InfraIn-bandIPMI/RedfishOn-PlatformOff-PlatformIn-bandIMA Devices,BMCIMA HS InfraOut-of-bandBMC DevicesBMC HS InfraIMA:In-band Management AgentIPMI:Intelligent Platform Management InterfaceMCTP:Management Component Tr

6、ansport ProtocolHS:Hyper ScalerIn-band/Out-of-band Management MethodsAuto-remediationSystem MonitoringIMA&BMC are complementarySystem-level AggregatorUnified modelHyperscale StorageHyper Scaler Management InfrastructureIn-bandOut-of-bandIn-band/Out-of-band Management MethodsHardware Fault Management