sijie commented on issue #1284: Improve FileInfoBackingCache URL: https://github.com/apache/bookkeeper/pull/1284#issuecomment-375752857 @ivankelly > it seems pointless to have locks around a concurrent hashmap well. depends on how you see stuffs. concurrent hashmap is section based implementation, it has better parallelism and less contention on accessing stuffs. especially most of the time the map is protected under a read lock and only be blocked when the map is going to modified. > I've made a PR into your branch with a lockless approach. I have thought about lockless approach but I didn't do it. A lockless approach moves the mutation out of the protected of a write lock, which can potentially make it exposed to race conditions that we don't think of. This area has been very tricky based on the PRs and discussions we had around it. It is a too-risky change for me. That's also the reason I explained above and the reason I don't feel comfortable about your lockless change. My immediate concern is only "don't do IO under one lock". I am much comfortable to keep the read/write lock until there is an evidence showing read/write lock is a problem.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
