Sessions are currently stored in Redis, a highly-optimized in-memory store with request latency reportedly on the order of ~1 ms. Given the nature of the proposed session storage service, latency can only be higher than what we're accustomed to; It seems implied that we are striking a bargain to trade away some latency in return for multi-master replication, but have not yet agreed upon a price.
We should to establish an SLA for the proposed session storage service.
Per discussion below:
|Action||Redis (ms)||Multiplier||Mean latency (ms)||p90 latency (ms)||error rate|