[#POOL-93] Reduce contention by making borrow & return more independent from each other

Commons Pool

Reduce contention by making borrow & return more independent from each other

Created: 24/Nov/06 03:23 PM Updated: 10/Dec/07 02:56 AM

Return to search

Component/s:

None

Affects Version/s:

1.1, 1.2, 1.3

Fix Version/s:

1.4

Time Tracking:

Not Specified

File Attachments:

File Name	Date Attached	Attached By	Size
GenericObjectPool.patch	2007-04-23 01:37 PM	Marcos Sanz	2 kB
GKOP-relaxedSyncOnReturn.patch	2006-11-24 03:24 PM	Holger Hoffstätte	4 kB
perf-patch-GenericPool.txt	2007-04-23 01:37 PM	Marcos Sanz	4 kB
pool-93-markt-v2.patch	2007-12-09 11:35 PM	Mark Thomas	27 kB
relaxedReturnObjectBenchmark.txt	2006-11-24 04:12 PM	Holger Hoffstätte	2 kB

Issue Links:

Blocker

This issue blocks:

~~DBCP-212~~ PoolingDataSource closes physical connections

Resolution Date:

10/Dec/07 02:56 AM

Description

« Hide

Currently borrow & return are completely blocked from each other, while at least the factory-based validation & destruction can be handled independently. A few simple changes narrow the synchronization blocks yet retain overall correctness.

All

Comments

Work Log

Change History

Subversion Commits

Sort Order:

Repository	Revision	Date	User	Message
ASF	#602773	Mon Dec 10 02:53:10 UTC 2007	psteitz	Reduced synchronization in GenericObjectPool, GenericKeyedObjectPool. Factory method activations within synchronized blocks were causing performance problems in DBCP and other applications where factory methods could block. JIRA: ~~POOL-93~~ JIRA: ~~POOL-108~~ Reported and patched by Holger Hoffstätte, Matthew Moore Patched by Marcos Sanz, Mark Thomas
				Files Changed
				MODIFY /commons/proper/pool/trunk/src/java/org/apache/commons/pool/impl/GenericKeyedObjectPool.java MODIFY /commons/proper/pool/trunk/src/java/org/apache/commons/pool/impl/GenericObjectPool.java MODIFY /commons/proper/pool/trunk/xdocs/changes.xml

Repository	Revision	Date	User	Message
ASF	#606295	Fri Dec 21 20:09:32 UTC 2007	psteitz	Ensure that _numActive is always incremented before starting makeObject and is decremented after destroyObject completes. Otherwise, while destroys are in progress, makes may kick off, resulting in too many concurrent instances. Test cases illustrate an unreported bug in pool 1.2, which does not affect 1.3. These changes should have been included in r602765. JIRA: ~~POOL-93~~ JIRA: ~~POOL-108~~
				Files Changed
				MODIFY /commons/proper/pool/branches/1_4_RELEASE_BRANCH/src/java/org/apache/commons/pool/impl/GenericKeyedObjectPool.java MODIFY /commons/proper/pool/branches/1_4_RELEASE_BRANCH/src/java/org/apache/commons/pool/impl/GenericObjectPool.java MODIFY /commons/proper/pool/branches/1_4_RELEASE_BRANCH/src/test/org/apache/commons/pool/impl/TestGenericObjectPool.java MODIFY /commons/proper/pool/branches/1_4_RELEASE_BRANCH/src/test/org/apache/commons/pool/impl/TestGenericKeyedObjectPool.java