Building scalable log streaming on FoundationDB

[ryanworl]

@begelundmuller

The documentation here states that range clears are “ultra-efficient”. But if I understand you correct, even range clears will consume significant resources after the fact to actually release the disk space?

Range clears incur no additional client-side latency. I wouldn’t call them efficient, though. They incur a debt that will be cleaned up asynchronously. There is a tipping point at which you won’t be able to clean pages fast enough, and your disk usage will grow faster than you can clean pages.