keith-turner commented on PR #5728: URL: https://github.com/apache/accumulo/pull/5728#issuecomment-3053042850
> I was wondering if we might be able to use a ZooCache instance here, instead of doing our own caching. Zoocache could help here. There are two things going on w/ this bug, there are N^2 reads from zookeeper and then there is N^2 CPU work done in the manager. If we used zoocache w/ the old code it would remove the N^2 reads from zookeeper, however the manager would still do N^2 CPU work. To reserve a single fate op the old manager code would sort all fate ops in memory and then skip over them to the last place it was. This change removes the N^2 reads and work in the manager. The new code is not caching in the same way as zoocache, because once it pulls things into memory threads start pulling things off the in memory queue until its empty at which point some thread will refill it. The refill of the queue in the new code could use zoocache to further reduce load on zookeeper. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: notifications-unsubscr...@accumulo.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
