[
https://issues.jboss.org/browse/ISPN-3102?page=com.atlassian.jira.plugin....
]
Adrian Nistor commented on ISPN-3102:
-------------------------------------
A little experiment, remove SKIP_INDEXING at put() and also remove rebuildIndexes() -- the
total time it takes to index at put is identical with the map-reduce based indexing. So at
least we know the slowness of the indexing is not due to a bug in map-reduce.
A second experiment, modify the test to run with ram directory provider or filesystem
provider, the test is super fast. We can conclude the performance problem is definitely
caused by the Infinispan directory provider. Tuning IndexWriter performance provided only
minor improvements.
For now I propose to lower the number of entries to 200 from 2000 as this is a functional
test, not a perf test. The performance of Infinispan directory provider should be
re-addressed.
Analyse the performance of MassIndexingTest
-------------------------------------------
Key: ISPN-3102
URL:
https://issues.jboss.org/browse/ISPN-3102
Project: Infinispan
Issue Type: Task
Reporter: Mircea Markus
Assignee: Adrian Nistor
Priority: Blocker
Fix For: 6.0.0.CR1
MassIndexingTest runs on CI in 10 mins with TRACE on and 3 mins with TRACE off. According
to Sanne this shouldn't take that long:
{quote}
sannegrinovero
11:50 mmarkus: the org.infinispan.query.distributed.MassIndexingTest test is adding 2000
entries to the cache. True that might be a bit excessive, still if it takes 10 minutes you
have a problem in Infinispan core.
mmarkus
11:52 sannegrinovero: each such write is indexed I take it?
sannegrinovero
11:53 mmarkus: no: withFlags(Flag.SKIP_INDEXING)
11:54 mmarkus: but it's a 4 node DIST test using Map/Reduce.
sannegrinovero
11:54 then the Map/Reduce job does some indexing but the indexing engine didn't
change overnight
{quote}
--
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