《OpenBMC port to RL300 a CSP - vendor perspective.pdf》由会员分享,可在线阅读,更多相关《OpenBMC port to RL300 a CSP - vendor perspective.pdf(13页珍藏版)》请在三个皮匠报告上搜索。
1、OpenBMC port to the RL300A CSP/Vendor perspectiveDeployments OpenBMC port to the RL300 A CSP/Vendor perspectiveElyes Zekri,Engineering Manager ScalewayJean-Marie Verdun,Distinguished Technologist GOEWho is Scaleway?A European cloud provider,founded in 1999.The most complete cloud ecosystem in Europe
2、Multi-AZ redundancy and native tools to architecture across multiple cloudsOperates sustainable data centers in France,the Netherlands and Poland100%powered by renewable energyDC5(PUE:1.15)uses free&adiabatic cooling only(no AC)Priority:minimize impact,maximize savingsThe cloud of choice:where data
3、is hosted,which ideal architecture.a more responsible way to scaleServers reused for up to 10 years(vs industry standard of 3-4):reduces CO2 and e-wasteHPE RL300 and OpenBMCOpenBMC current state on GXPCortex A9L2 cacheMemory controllerDDR3/DDR416AHB BusAXI to AHBI2C engine10/100/1000 MACCIFSPI engin
4、es(x3)Interrupt controller(s)UART(s)VideoPCI RegistersLPCPECIeSPI/LPCPECIPCIe x1OpenBMC current state on RL300Basic features are here!Fan/Cooling ManagementInventory managementKVM/SoLPower Management(PSU monitoring and reporting)Error monitoring(including CPU/DIMMs)Plenty of sensorsIPMI/Redfish SELS
5、till lacking but under developmentFront I/O(nvme monitoring)PCIe firmware updateOpenBMC and ScalewayScaleway is maintaining a huge number of old and new servers managed by several FW stacks from different suppliers need the most common possible HW management interfaceAccess to energy data is becomin
6、g very challenging but power consumption and cooling data is not always exposed as expected need to have a customizable way to control and monitor power and cooling at the chassis level from the FWOpenBMC could be THE appropriate answer to address these needsScaleway-HPE collaborationHPE provided a
7、proactive day-to-day support to Scaleway and shared all resources required to run OpenBMC on RL300:OpenBMC source tree including RL300 build target and Network boot layer documentation regarding:OpenBMC build procedureILO-to-OpenBMC transition proceduretransfer of ownership procedureFW Network boot
8、enabling procedure(need of a debug serial cable)common ILO/OpenBMC host ROM flashing Scaleway OpenBMC PoC(work in progress)PoC Framework:Hardware pre-production validation and testingUsing OpenBMC APIs to update/configure/check BMC and ROM FW and to perform hardware inventory What has been doneScale
9、way is now able to perform code changes and rapidly test them with FW network bootEth-over-usb interface has been enabled to allow host-BMC interactionProprietary HPE/ILOREST management commands are being translated to standard Redfish commands What might be next for ScalewayUsing OpenBMC as a singl
10、e multi-vendor FW stack may be a good opportunity to:maintain one single image in our own ci-cd environmentimplement our own custom FW features when not fully satisfied by the supplier implementation rely on a secure/interoperable/standard management API based on Redfishbreak the dependency with the
11、 various vendors and suppliers(especially when the chassis/servers become EOS)OpenBMC will help us to improve our velocity to go to productionWhat might be next for ProLiantFaster upstreamingNetwork boot hardeningAuto updateLocal firmware managementZero Trust firmware approachFRU Management is a cha
12、llenge,need industry standard Testing is key,especially in production environment,OSF opens new ways to manage huge fleet of servers,we need to explore them and share outcome.OEM provides a good abstraction layer to chips vendors ecosystem which might be tricky to handle for consumers into higher level within the value chain.OpenBMC on ProLiant Gen11 enablement will be released in multiple steps(April to June 2023)Testing can be done through OSFCI,an HPE contribution to OCP available at http:/ to Action Thank you!