dyf6372 opened a new pull request #6719: use SnapTreeMap instead of ConcurrentSkipListMap URL: https://github.com/apache/incubator-druid/pull/6719 Using jprofiler, I found that the function in ConcurrentSkipListMap was cost more than 30% cpu in our case.  SnapTreeMap is a drop-in replacement for ConcurrentSkipListMap. Form the paper http://ppl.stanford.edu/papers/ppopp207-bronson.pdf： _Experimental evidence shows that our algorithm outperforms a highly tuned concurrent skip list for many access patterns, with an average of 39% higher singlethreaded throughput and 32% higher multi-threaded throughput over a range of contention levels and operation mixes._ After I changing to SnapTreeMap , it's 20%~30% faster than ConcurrentSkipListMap. 
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
