apurtell edited a comment on issue #1407: HBASE-24099 Use a fair ReentrantReadWriteLock for region close lock URL: https://github.com/apache/hbase/pull/1407#issuecomment-610102467 Digging into scan degradation with PE, let me say first it took me a while to find a good test case with PE. Wanted to avoid contention on row locks and yet hit region locks. Finally I wrote a table with 100M rows, then varied parameters so for each case no matter how many threads (via --nomapred) the full range of rows would be considered when randomly picking start keys, and 100000 operations would be attempted. It looks like point reads are between 9-15% faster, and scans are degraded from 2-14%. scanRange10 is 20% faster, was repeatable, and is an unexplained anomaly. This was a cluster with one master, five regionservers, on AWS M5 class instances, Hadoop 2.10.0, HBase 1.7.0-SNAPSHOT, PE on a separate client box. [pe_results.pdf](https://github.com/apache/hbase/files/4441386/pe_results.pdf)
---------------------------------------------------------------- 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: us...@infra.apache.org With regards, Apache Git Services
