[ https://issues.apache.org/jira/browse/CASSANDRA-3762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13192654#comment-13192654 ]
Pavel Yaskevich commented on CASSANDRA-3762: -------------------------------------------- I mention this because the problem in the original ticket was with rolling restarts taking too much time on index summary computation (read going though whole PrimaryIndex for every SSTable out there), so imagine situation when you have few hundreds of SSTables each with key cache in the different parts of the primary index this means if you go with getPosition() calls you will have a lot of random I/O (meaning you will have to seek deeper and deeper into the primary index file which means slower data access even in mmap mode) on each of those and I'm not sure if it's really better than reading primary index sequentially especially knowing that you have already read all of the index/data positions from the Summary component. I propose you do the test with many SSTables and compare system load times (don't forget to drop page cache between tests with `sync; echo 3 > /proc/sys/vm/drop_caches`). By the way, I forgot to ask you if you dropped page cache before running second test? if you didn't that would pretty much explain such a dramatic improvement in the load time... > AutoSaving KeyCache and System load time improvements. > ------------------------------------------------------ > > Key: CASSANDRA-3762 > URL: https://issues.apache.org/jira/browse/CASSANDRA-3762 > Project: Cassandra > Issue Type: Improvement > Components: Core > Affects Versions: 1.2 > Reporter: Vijay > Assignee: Vijay > Priority: Minor > Fix For: 1.2 > > Attachments: 0001-SavedKeyCache-load-time-improvements.patch > > > CASSANDRA-2392 saves the index summary to the disk... but when we have saved > cache we will still scan through the index to get the data out. > We might be able to separate this from SSTR.load and let it load the index > summary, once all the SST's are loaded we might be able to check the > bloomfilter and do a random IO on fewer Index's to populate the KeyCache. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira