[
https://issues.apache.org/jira/browse/HBASE-9272?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13747277#comment-13747277
]
Lars Hofhansl commented on HBASE-9272:
--------------------------------------
I was thinking about this too (i.e. keeping all RSs busy). On the other hand I
was trying to keep this simple, assuming that in most cases the region to
server assignment would be more or less random.
With some number of threads and a reasonable sized cluster (without which a
parallel scanner does not help much anyway), one would assume a fairly nice
load distribution.
So a test with more regions should see the same speedup, there is nothing
inherently costly per region (the ClientScanners will need to find the region
again, but it should be cached).
There are other considerations too. For example, instead of having a task per
Region, one could split the requested rowkey space into N slices (using the
region boundaries as a poor-mans histogram by assuming that all regions will be
of roughly the same size in bytes). In that case one would keep the number of
threads unlimited but instead limit the number of tasks (i.e. slices).
(also above the penalty was 2.2% rather than 1.5% - but that was just a single
run anyway)
Will do a test with more regions.
> A simple parallel, unordered scanner
> ------------------------------------
>
> Key: HBASE-9272
> URL: https://issues.apache.org/jira/browse/HBASE-9272
> Project: HBase
> Issue Type: New Feature
> Reporter: Lars Hofhansl
> Assignee: Lars Hofhansl
> Priority: Minor
> Attachments: ParallelClientScanner.java, ParallelClientScanner.java
>
>
> The contract of ClientScanner is to return rows in sort order. That limits
> the order in which region can be scanned.
> I propose a simple ParallelScanner that does not have this requirement and
> queries regions in parallel, return whatever gets returned first.
> This is generally useful for scans that filter a lot of data on the server,
> or in cases where the client can very quickly react to the returned data.
> I have a simple prototype (doesn't do error handling right, and might be a
> bit heavy on the synchronization side - it used a BlockingQueue to hand data
> between the client using the scanner and the threads doing the scanning, it
> also could potentially starve some scanners long enugh to time out at the
> server).
> On the plus side, it's only a 130 lines of code. :)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
