Closed willscott closed 7 years ago
Just 1 comment, not sure I understand why we're padding batchreadrequest to the same size.
Okay to leave readbatch padding at centralized for now, and push the logic into the shard when we refactor frontend/centralized/shard responsibilities?
👍
when testing, we'll often have a single client and a single read at a given them. Centralize will now fill up the batch of reads with additional empty requests every
ReadInterval
so that the read will always complete in a timely manner in these cases.