We have had many customers ask for Redfish on our servers' BMCs. But recently a customer asked for Redfish-like functionality on our Host CPU. Not just the static CPU, DRAM, PCIe, NIC, etc. data that comes from BIOS start, but real time interactive view and configuration of things like storage (e.g., RAID) modes and NIC configuration like IPv4 and IPv6 addresses, etc.
This strikes me as a logical extension of the Redfish paradigm. I am, of course, aware of Swordfish, which clearly does most or all of the storage related things I can imagine someone needing. But what about networking? What about other OS configuration and reporting?
Does anyone know of Redfish REST API driven projects in the world to do these things or of demand for such? Does anyone have an opinion (hah) on whether something like this is a good idea?
Alan Mimms System Architect Jabil Enterprise and Infrastructure division
Does the Host CPU have any local attached storage ? Or is it just a compute node with storage being provided from an external dedicated storage array?
If there is no storage functionality being provided by the Host CPU, I am not sure whether any storage profile wrt Swordfish should be supported. In this case, the storage entity which is connected to the Host CPU shall implement the HOSTED SERVICE config related to Swordfish and provide the required storage functionality. The Host CPU then would become a IO Endpoint for the Storage array.
Effectively, both the Storage Entity and the Host CPU should provide diagnostic/ config APIs per Redfish/ Swordfish spec and the centralised Pod Manager which has access to both the Host CPU and the Storage Array should be responsible for overall aggregation/ orchestration.