Memorystore for memcached

[GregF]

Appengine python2.7 standard environment. We're looking at switching from the built-in memcache to memorystore for memcached (M4M), as a precursor to migrating to Python3.

1. The M4M pricing doc (https://cloud.google.com/memorystore/docs/memcached/pricing) describes vCPUs, but gives go guidance on throughput. Does anyone have figures for how many memcache requests/sec one vCPU can handle, and would two vCPUs handle twice that?

2. You need a VPC instance to connect Appengine to your M4M instance. Again, there is no guidance for throughput - how many requests can a VPC vCPU handle, and how much memory should we provision it with?

3. Appengine standard has no explicit region, but you have to specify a region for your M4M. What region should we pick?

Thanks in advance for any advice.