Atomikos Forum

Important question to the community

Hi all,

One of the features in our backlog is: optimizing for 1-phase commit (i.e., when only one resource is involved in the transaction).

However, the killer use case for Atomikos is precisely when 2 or more resources are needed. Consequently, we've always been reluctant to changing the design of our core algorithms to fully support 1-phase commit in all possible cases.

The question to you: what do you think? Should we invest in redesigning our core to go for 1-phase commit even when we think this is marginal?

Guy Pardon Send private email
Thursday, March 22, 2012
Would these optimization only apply when multiple datasources are being used and only one is modified?  Or would it benefit single datasource non xa situations as well?

I'm inclined to say don't change your algorithm.
Mike Youngstrom Send private email
Sunday, April 01, 2012
It would only benefit in the following case:

-only one resource (DataSource/ConnectionFactory) is used in the transaction, AND

-at least one Synchronization is registered

In practice, this translates into:

-Using JTA with Hibernate and only having JDBC (no JMS) to one database.

Guy Pardon Send private email
Monday, April 02, 2012
I'll let you make the call but that is my use case 99% of the time.  So, a boost in that use case may be useful to me.
Mike Youngstrom Send private email
Monday, April 09, 2012
Thanks Mike!

I am a bit surprised that JTA is a killer use case for Hibernate just by itself; do you mind explaining the rationale behind that?

Guy Pardon Send private email
Wednesday, April 11, 2012
James Depoorter Send private email
Tuesday, April 24, 2012

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics