[
https://issues.jboss.org/browse/ISPN-3891?page=com.atlassian.jira.plugin....
]
Dan Berindei commented on ISPN-3891:
------------------------------------
Oops, I was thinking of a {{cache.get(k)}} call without any running tx, but I forgot that
doesn't create a tx at all. Still, when recovery is not enabled, we can probably avoid
enlisting with the TM even for implicit txs created by write operations.
When doing a get within a running tx, I think the only thing that should matter is whether
the FORCE_WRITE_LOCK flag was used (because then you're actually doing a
{{lockAndReturnCurrentValue(k)}}). Note that I'm not proposing to change anything
else, and the value should still be stored in the invocation context to be reused by other
operations in the same tx/batch.
Tune for batching without transactions
--------------------------------------
Key: ISPN-3891
URL:
https://issues.jboss.org/browse/ISPN-3891
Project: Infinispan
Issue Type: Enhancement
Components: Transactions
Reporter: Sanne Grinovero
Assignee: Mircea Markus
Fix For: 7.0.0.Beta1, 7.0.0.Final
The usage of batching is a critically important feature, widely used to improve
performance in many common scenarios, but while documentation makes it clear that it's
built on transactions, it's unclear how these transactions should be configured.
Also, I would challenge the decision of building batching on top of transactions and see
if there are opportunities for further optimization.
--
This message was sent by Atlassian JIRA
(v6.2.3#6260)