On 24 Jan 2011, at 16:34, Sanne Grinovero wrote: The performance difference would result from TM doing write on the disk during 2PC prepare( I think it's after the prepare phase). This only happens if we run distributed transactions tough (i.e. more than one participant), otherwise the TM would optimise the call for a single resource and no disk access takes place. Even more, AFAIK this can be disabled not to do disk access ( keep tx log in memory). So I estimate the performance should be about the same.

On 11-01-25 2:12 PM, Manik Surtani wrote: +1

I've reverted Hudson and added a JIRA for enabling JBossTM: https://issues.jboss.org/browse/ISPN-898 On 25 Jan 2011, at 19:04, Manik Surtani wrote:

On 26 Jan 2011, at 12:17, Mircea Markus wrote: It should be quicker. There is no maintenance of a persistent tx log. Nope. I agree with you that there is no point expending effort to enhance the DummyTM - but IMO I don't think this is necessary right now. Cheers Manik -- Manik Surtani manik(a)jboss.org twitter.com/maniksurtani Lead, Infinispan http://www.infinispan.org

On 26 Jan 2011, at 12:30, Mircea Markus wrote: There are a few tests testing transactions that span different cache instances. Each one would be considered as its own RM. Not true. https://github.com/infinispan/infinispan/blob/master/core/src/main/java/o... Cheers Manik -- Manik Surtani manik(a)jboss.org twitter.com/maniksurtani Lead, Infinispan http://www.infinispan.org

On 26 Jan 2011, at 13:08, Mircea Markus wrote: Yeah good point. -- Manik Surtani manik(a)jboss.org twitter.com/maniksurtani Lead, Infinispan http://www.infinispan.org

+1! Erik From: infinispan-dev-bounces(a)lists.jboss.org [mailto:infinispan-dev-bounces@lists.jboss.org] On Behalf Of Mircea Markus Sent: Monday, January 24, 2011 10:35 AM To: infinispan -Dev List Subject: [infinispan-dev] running test suite with JBossTM vs DummyTM Hi, Right now we are running our test suite using DummyTM. This is mainly for suppling users with a default, lightweight TM. And avoid a additional dependency. Even though this worked well for us so far, this might not be such a good idea after all: there are many not-so-obvious rules to be considered when implementing such a transaction manager which, if not implemented might make our XAResource implementation work badly even though it works fine(eg [1]). Or even worse, it might make it look okay even though it doesn't. On top of that there are grey areas in JTA spec, for which different vendors use different approaches. By using/testing against JBossTM we can document these usages and stick with (the recommended?) JBossTM functionality. Taking this one step further, wondering weather it makes sense to keep the DummyTM as it is in our code base, given the fact that it is far away from implementing the spec. One solution would be to enhance it (-1 IMO), another to replace it. Perhaps with a lightweight JBossTM? Cheers, Mircea [1] http://lists.jboss.org/pipermail/infinispan-dev/2011-January/007243.html ________________________________ The information contained in this message is legally privileged and confidential, and is intended for the individual or entity to whom it is addressed (or their designee). If this message is read by anyone other than the intended recipient, please be advised that distribution of this message, in any form, is strictly prohibited. If you have received this message in error, please notify the sender immediately and delete or destroy all copies of this message.