hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kannan Muthukkaruppan (JIRA)" <j...@apache.org>
Subject [jira] [Created] (HBASE-4466) Fix accounting of the number of blocks read for the one-level index case
Date Fri, 23 Sep 2011 05:01:27 GMT
Fix accounting of the number of blocks read for the one-level index case
------------------------------------------------------------------------

                 Key: HBASE-4466
                 URL: https://issues.apache.org/jira/browse/HBASE-4466
             Project: HBase
          Issue Type: Improvement
            Reporter: Kannan Muthukkaruppan
            Assignee: Mikhail Bautin


In HFileBlockIndex.seekToDataBlock if the current block is the same as the requested block,
then in the 1-level index case, we were read block from cache again. Although this would be
a cache hit, it unnecessarily increases the block cache read count and block cache hit count,
and also makes it harder to keep the accounting straight for tests like TestBlocksRead (introduced
in HBASE-4450).

Basically, even in the 1-level index case, in HFileBlockIndex.seekToDataBlock(), if currentBlock.getOffset()
== currentOffset, we can avoid looking up the block in cache. 

Assigning to Mikhail (he's already fixed this in our internal branch).



--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message