GitHub user ramkrish86 opened a pull request:
https://github.com/apache/phoenix/pull/12
Phoenix-180
A fresh pull request with all the changes for guideposts.
Tested in a single cluster for different combinations.
"SELECT count(*) FROM %s WHERE host='CS' AND domain='Apple.com' AND
FEATURE >= 'Dashboard' AND FEATURE <= 'Report';" % (table))
With 5000000 rows
Withoutpatch
===========
COUNT(1)
----------
555169
Time: 2.264 sec(s)
Withpatch
=========
COUNT(1)
----------
555169
Time: 0.813 sec(s)
Withoutpatch
==========
COUNT(1)
----------
1002201
Time: 1.889 sec(s)
Withpatch
==========
COUNT(1)
----------
1002201
Time: 1.075 sec(s)
Am sure we would get better results if we are able to tune the guidePosts
width so that instead of fetching in the range of 1000s inside a region we can
fetch 10ks of result in every chunk. Also should work with bigger tables and
more rows.
The tests were performed with 256*1024 and 1024 * 1024 as the
guidepostsWidth size.
The main thing you can observe in the above case is that the query
targetted to single region is parallelized and distributed to smaller chunks
and so the work done by every thread is less.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/ramkrish86/phoenix-1 stats_new
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/phoenix/pull/12.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #12
----
commit b1950f9363b655ec2bd5a4263e29360958d5dc38
Author: Ramkrishna <[email protected]>
Date: 2014-09-10T13:03:21Z
Phoenix-180
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
