[infinispan-dev] An observation about total order and consistency

[Paolo Romano]

Good, this means that those results were based on a fair comparison!

     Paolo

On 3/7/12 6:49 PM, Pedro Ruivo wrote:
> BTW, the results posted, both configurations had the flag
> syncCommit/RollbackPhase set to false.
>
> On 3/7/12 1:41 PM, Paolo Romano wrote:
>> I did not know about this change, thanks for pointing it out! By the
>> way, is it still possible to send commit messages asynchronously with
>> 2PC by changing the default config? I am asking as, in order to ensure
>> fairness when comparing the performance of the two protocols, it would
>> be better to make sure that they enforce the same semantic.
>>
>> Another possibility would be to extend the TOB-based replication
>> mechanism to support this semantic. Conceptually, this is not too
>> complicated, but coding that will still take some time... and I  would
>> prefer to invest effort on TOM+partial replication. What do you think?
>>
>> Cheers,
>>
>>        Paolo
>>
>> On 3/7/12 1:17 PM, Mircea Markus wrote:
>>>> if I recall correctly currently (with 2PC)  the commit is sent
>>>> asynchronously (without waiting to gather acks).
>>> The commit used to be send async (default) before 5.2, but now it is sent sync.
>>> *But* there were no complains with async commits AFAIK, which is a good thing.
>>>
>>>
>>> _______________________________________________
>>> infinispan-dev mailing list
>>> infinispan-dev at lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>> _______________________________________________
>> infinispan-dev mailing list
>> infinispan-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev