Outcome

RHOSO currently doesn't have possibility to effectively manage resources in dynamic fashion.  

This outcome aims to implement and (or) integrate (existing) solution of dynamic management tool or framework for resources such as GPUs, NVMe devices or MIG slices. The tool/framework needs be able to provide the managed resources to Nova service, so that it is possible to attach such devices to VM instances in Nova.

This feature will be beneficial for OpenStack users, who won't need to manage accelerator resources manually using low level configuration of HW and OpenStack services.
 
Success Criteria

It is possible to manage accelerator resources such as GPUs,  NVMe resources or MIG slices via OpenStack user interface (whether using CLI or GUI).

Expected Results (what, how, when)

Dynamic management of accelerator resources is key feature for AI workloads which improves competitiveness of the RHOSO product from alternatives available on the market. This work is planned to be delivered in RHOSO-19.0 timeframe.

Post Completion Review – Actual Results

After completing the work (as determined by the "when" in Expected Results above), list the actual results observed / measured during Post Completion review(s).