FY22-Q1 KR: Get signed-off DB sharding architecture evolution => 50%

CC @sloyd

Retrospective

Good

  1. Landed on a POC prototyping plan (horizontal top-level namespace sharding) and a secondary backup plan (vertical application sharding).
  2. Assembled a dedicated sharding group.
  3. Kicked off the POC - spikes, researches, and prototyping.
  4. Improved the WG meetings after receiving feedback.

Bad

  1. The sharding approach proposals came together in April (later half of the quarter), due to the delayed WG kickoff because of other burning priorities of the DRI and the fact that sharding was not in scope initially when establishing the Database Scalability Working Group.
  2. The WG meetings didn't go well in the first month from biased for actions and laying out ETA perspectives.
  3. It was challenging to assemble the sharding group.

Try

  1. More quantifiable information, such as confidence level, impact level (to codebase and functional domains), progress %, ETA etc., although the POC is nebulous in nature.
  2. Always bias for actions and take timely actions when DRI becomes bottleneck.
Edited by Eric Johnson