[infinispan-dev] Performance improvements, more...

[Bela Ban]

It would be interesting to see the numbers with bbc128, which makes 
sending a bit faster. I'd expect to see more writes and less reads, 
compared to their relative numbers.

BTW: I'm done with my implementation of Table, and the numbers look 
really impressive ! It is about the same as RingBuffer for smaller 
insertions (5 million), but for 50 million the number stays about the 
same (insertions and removals per second). For smaller numbers, Table is 
ca 4 times *faster* than NakReceiverWindow.

I still want to add more tests for Table (copy and convert the ones for 
RingBuffer), and then switch NAKACK2 over from RingBuffer to Table. I'm 
very curious to see the perf numbers after that change !

Next comes passing up of entire bundles, this should also make a big 
difference !
Exiting times, cheers !

On 1/19/12 10:50 AM, Sanne Grinovero wrote:
> On 19 January 2012 09:36, Bela Ban<bban at redhat.com>  wrote:
>> Great !
>>
>> Which version of JGroups did you use ? 3.0.2 or 3.0.3 ?
>
> Right... sorry I forgot to revert experiments on JGroups!
> So both numbers where running JGroups "version" 8115f27, which is
> somewhere between the two:
>
> 8115f27 Made MessageDispatcher.members volatile, writes and reads are
> separated by memo barriers (JGRP-1414)
> cb6219d Changes to Rsp (https://issues.jboss.org/browse/JGRP-1412)
> 070b116 updated version to 3.0.3.Final
> 8f485fa (tag: JGroups_3_0_2_Final) changed version to 3.0.2.Final
>
>
>> On 1/19/12 10:25 AM, Sanne Grinovero wrote:
>>> Hello,
>>> I just got these figures using the Transactional test:
>>>
>>> Done 2,592,102,434 transactional operations in 62.18 minutes using
>>> 5.1.0-SNAPSHOT
>>> Â  Â 2,576,372,975 reads and 15,729,459 writes
>>> Â  Â Reads / second: 690,573
>>> Â  Â Writes/ second: 4,216
>>>
>>> which are much better than what I had 24 hours ago with 23a031e
>>> (same test, same parameters) :
>>>
>>> Done 878,390,104 transactional operations in 23.06 minutes using 5.1.0-SNAPSHOT
>>> Â  Â 873,278,200 reads and 5,111,904 writes
>>> Â  Â Reads / second: 631,100
>>> Â  Â Writes/ second: 3,694
>>>
>>> The main differences are our minor cleanups from yesterday and
>>> JBMAR-127 - but this wasn't released yet, I'm depending on a locally
>>> built snapshot.
>>>


-- 
Bela Ban
Lead JGroups (http://www.jgroups.org)
JBoss / Red Hat