[JBoss JIRA] Created: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
6:42 p.m.
Cache performance slow in a JTA transaction with many cache operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Bug
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
When running Infinispan in a JTA transaction the performance of cache operations decreases
as the number of operations increases in the transaction. If adding to the cache, adding
50k items is much faster when done 1k items per transaction than all 50k in one
transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for the
all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
6:51 p.m.
New subject: [JBoss JIRA] Commented: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Tom Waterhouse commented on ISPN-1047:
--------------------------------------
The two tests that best illustrate the issue are TXCacheTest.testWithManyBatches() and
TXCacheTest.testWithOneBatch(). Both do the exact same operations, but
testWithManyBatches() does the cache operations in 50 1k batches, while testWithOneBatch()
does 1 50k batch.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Bug
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
6:54 p.m.
New subject: [JBoss JIRA] Updated: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Tom Waterhouse updated ISPN-1047:
---------------------------------
Attachment: ISPN-1047.zip
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Bug
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
6:44 a.m.
New subject: [JBoss JIRA] Updated: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Manik Surtani updated ISPN-1047:
--------------------------------
Issue Type: Quality Risk (was: Bug)
Labels: performance (was: )
Assignee: Mircea Markus (was: Manik Surtani)
Fix Version/s: 4.2.2.BETA1
Priority: Critical (was: Major)
Complexity: Medium
Component/s: Transactions
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Mircea Markus
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
7 a.m.
New subject: [JBoss JIRA] Updated: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Mircea Markus updated ISPN-1047:
--------------------------------
Fix Version/s: 5.0.0.FINAL
(was: 4.2.2.BETA1)
Not sure this is a candidate for 4.2.2. Transactions are heavily revisited in 5.0: not
only enhancements, bug fixes fixes as well.
Performance should be looked at post 5.0 stuff.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Mircea Markus
Priority: Critical
Labels: performance
Fix For: 5.0.0.FINAL
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:17 p.m.
New subject: [JBoss JIRA] Updated: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Manik Surtani updated ISPN-1047:
--------------------------------
Assignee: Manik Surtani (was: Mircea Markus)
Fix Version/s: 4.2.2.BETA1
(was: 5.0.0.FINAL)
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:19 p.m.
New subject: [JBoss JIRA] Commented: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Manik Surtani commented on ISPN-1047:
-------------------------------------
This has nothing to do with the XA impl, it has to do with the way read committed
visibility guarantees are maintained in the contexts.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:19 p.m.
New subject: [JBoss JIRA] Issue Comment Edited: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Manik Surtani edited comment on ISPN-1047 at 4/18/11 1:18 PM:
--------------------------------------------------------------
This has nothing to do with the XA impl, it has to do with the way read committed
visibility guarantees are maintained in the contexts.
In response to Mircea's comment, transactions have been revisited in 5.0 but this
section (read visibilities and locking) remains the same. So the bug is still relevant
and exists in both 4.2.x and 5.x
was (Author: manik):
This has nothing to do with the XA impl, it has to do with the way read committed
visibility guarantees are maintained in the contexts.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:34 p.m.
New subject: [JBoss JIRA] Resolved: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Manik Surtani resolved ISPN-1047.
---------------------------------
Resolution: Done
This is fixed. Pushing a 4.2.2-SNAPSHOT out now for you to try out.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:40 p.m.
New subject: [JBoss JIRA] Commented: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Tom Waterhouse commented on ISPN-1047:
--------------------------------------
Outstanding. I'll let you know my test results as soon as I am able to access the
build.
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:50 p.m.
New subject: [JBoss JIRA] Commented: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Tom Waterhouse commented on ISPN-1047:
--------------------------------------
The results look the similar for adding with one batch and with many smaller batches.
One batch:
[2011-47-18 10:47:14.44, 1357]INFO [main](TXCacheTest.java:81) - starting
testWithOneBatch
[2011-47-18 10:47:14.71, 1384]WARN [main](arjunaI18NLoggerImpl.java:523) - ARJUNA-12130
Name of XA node not defined. Using Arjuna:1293
[2011-47-18 10:47:16.624, 3937]INFO [main](TXCacheTest.java:90) - testWithOneBatch
complete - 2579ms <----
Many batches:
[2011-47-18 10:47:41.317, 3909]INFO [main](TXCacheTest.java:71) - 48000 test data batch
added to cache
[2011-47-18 10:47:41.321, 3913]INFO [main](TXCacheTest.java:71) - 49000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:71) - 50000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:76) - testWithManyBatches
complete - 2571ms <----
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:50 p.m.
New subject: [JBoss JIRA] Issue Comment Edited: (ISPN-1047) Cache performance slow in a JTA transaction with many cache operations
[
https://issues.jboss.org/browse/ISPN-1047?page=com.atlassian.jira.plugin....
]
Tom Waterhouse edited comment on ISPN-1047 at 4/18/11 1:49 PM:
---------------------------------------------------------------
The results look similar for adding with one batch and with many smaller batches.
One batch:
[2011-47-18 10:47:14.44, 1357]INFO [main](TXCacheTest.java:81) - starting
testWithOneBatch
[2011-47-18 10:47:14.71, 1384]WARN [main](arjunaI18NLoggerImpl.java:523) - ARJUNA-12130
Name of XA node not defined. Using Arjuna:1293
[2011-47-18 10:47:16.624, 3937]INFO [main](TXCacheTest.java:90) - testWithOneBatch
complete - 2579ms <----
Many batches:
[2011-47-18 10:47:41.317, 3909]INFO [main](TXCacheTest.java:71) - 48000 test data batch
added to cache
[2011-47-18 10:47:41.321, 3913]INFO [main](TXCacheTest.java:71) - 49000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:71) - 50000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:76) - testWithManyBatches
complete - 2571ms <----
was (Author: tomwaterhouse):
The results look the similar for adding with one batch and with many smaller batches.
One batch:
[2011-47-18 10:47:14.44, 1357]INFO [main](TXCacheTest.java:81) - starting
testWithOneBatch
[2011-47-18 10:47:14.71, 1384]WARN [main](arjunaI18NLoggerImpl.java:523) - ARJUNA-12130
Name of XA node not defined. Using Arjuna:1293
[2011-47-18 10:47:16.624, 3937]INFO [main](TXCacheTest.java:90) - testWithOneBatch
complete - 2579ms <----
Many batches:
[2011-47-18 10:47:41.317, 3909]INFO [main](TXCacheTest.java:71) - 48000 test data batch
added to cache
[2011-47-18 10:47:41.321, 3913]INFO [main](TXCacheTest.java:71) - 49000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:71) - 50000 test data batch
added to cache
[2011-47-18 10:47:41.324, 3916]INFO [main](TXCacheTest.java:76) - testWithManyBatches
complete - 2571ms <----
Cache performance slow in a JTA transaction with many cache
operations
----------------------------------------------------------------------
Key: ISPN-1047
URL: https://issues.jboss.org/browse/ISPN-1047
Project: Infinispan
Issue Type: Quality Risk
Components: Transactions
Affects Versions: 4.2.1.FINAL
Environment: Infinispan 4.2.1.FINAL/Hibernate 3.6.1.Final/Spring 3.0.5/JBossJTA
4.14.0
Reporter: Tom Waterhouse
Assignee: Manik Surtani
Priority: Critical
Labels: performance
Fix For: 4.2.2.BETA1
Attachments: ISPN-1047.zip
When running Infinispan in a JTA transaction the performance of cache operations
decreases as the number of operations increases in the transaction. If adding to the
cache, adding 50k items is much faster when done 1k items per transaction than all 50k in
one transaction.
Using JProfiler the bulk of the time was spent in LockingInterceptor.doAfterCall() for
the all 50k in one transaction test.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
4997
days inactive
5001
days old
infinispan-issues@lists.jboss.org
11 comments
3 participants
participants (3)
-
Manik Surtani (JIRA) -
Mircea Markus (JIRA)
-
Tom Waterhouse (JIRA)