apurtell edited a comment on pull request #3244: URL: https://github.com/apache/hbase/pull/3244#issuecomment-841533295
> Thanks for the detailed perf results, SNAPPY seems like the sweet spot. Avg append time took a good hit? (10x IIUC?), something to be concerned about? The extra latency is going to be very much workload dependent, and so WAL value compression should remain a non-default configuration option. The test case I used to produce this benchmark employs very large values. We are loading in entries from Common Crawl -- web pages, images, javascript, CSS, big xml. So, that is both the source of the good compression results -- the values are large and compressible, generally -- and the source of the extra WAL append latency -- the values are large, generally, so the compression codecs have a fair amount of work. Even so, SNAPPY only adding ~2ms per append under these circumstances is a great result. For smaller values the latency hit will be proportionally lesser. > Btw, is this ready for review or are you still iterating on the code? See the findbugs warning above and just assume I will fix that. Otherwise I'm waiting for further feedback. @bharathv -- 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. For queries about this service, please contact Infrastructure at: [email protected]
