vNUMA (virtual NUMA) is a memory optimisation technology that makes virtual machine aware of the NUMA topology of the underlying physical server topology, which is very important for some specific kind of workload like HPC. Some significant work on vNUMA on PV has been done by Elena Ufimtseva, vNUMA on HVM was also posted by other developers, but none of them was merged upstreamed. This talk will cover the history, design and implementation of vNUMA, and possibly with some number to back up the importance of this feature.
