[
https://issues.apache.org/jira/browse/HBASE-5979?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13280640#comment-13280640
]
Todd Lipcon commented on HBASE-5979:
------------------------------------
Hey Kannan,
Sorry, let me elaborate on that suggestion:
The idea is to make a new FSReader implementation, which only has one API. That
API would look like the current positional read call (i.e take a position and
length).
Internally, it would have a pool of cached DFSInputStreams, and remember the
position for each of them. Each of the input streams would be referencing the
same file. When a read request comes in, it is matched against the pooled
streams: if it is within N bytes forward from the current position of one of
the streams, then a seek and read would be issued, synchronized on that stream.
Otherwise, any random stream would be chosen and a position read would be
chosen. Separately, we can track the last N positional reads: if we detect a
sequential pattern in the position reads, we can take one of the pooled input
streams and seek to the next predicted offset, so that future reads get the
sequential benefit.
> Non-pread DFSInputStreams should be associated with scanners, not
> HFile.Readers
> -------------------------------------------------------------------------------
>
> Key: HBASE-5979
> URL: https://issues.apache.org/jira/browse/HBASE-5979
> Project: HBase
> Issue Type: Improvement
> Components: performance, regionserver
> Reporter: Todd Lipcon
>
> Currently, every HFile.Reader has a single DFSInputStream, which it uses to
> service all gets and scans. For gets, we use the positional read API (aka
> "pread") and for scans we use a synchronized block to seek, then read. The
> advantage of pread is that it doesn't hold any locks, so multiple gets can
> proceed at the same time. The advantage of seek+read for scans is that the
> datanode starts to send the entire rest of the HDFS block, rather than just
> the single hfile block necessary. So, in a single thread, pread is faster for
> gets, and seek+read is faster for scans since you get a strong pipelining
> effect.
> However, in a multi-threaded case where there are multiple scans (including
> scans which are actually part of compactions), the seek+read strategy falls
> apart, since only one scanner may be reading at a time. Additionally, a large
> amount of wasted IO is generated on the datanode side, and we get none of the
> earlier-mentioned advantages.
> In one test, I switched scans to always use pread, and saw a 5x improvement
> in throughput of the YCSB scan-only workload, since it previously was
> completely blocked by contention on the DFSIS lock.
--
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
