[infinispan-dev] benchmarking ideal distribution

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[infinispan-dev] benchmarking ideal distribution

Mircea Markus
Hi,


Cheers,
Mircea

_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|

Re: [infinispan-dev] benchmarking ideal distribution

Galder Zamarreno
Some comments, in:

name="idealdistribution/dist-sync-ideal-distribution.xml" useEvenHashing="true

Does dist-sync-ideal-distribution.xml differ in any way to dist-sync.xml? The reason I ask is cos reading the wiki, it seems like useEvenHashing "makes Infinispan plugin to use the ideal consistent hashing implementation rather than the configured one". So, if dist-sync-ideal-distribution.xml and dist-sync.xml are different, in which way are they different and way? Seems like RadarGun could already make the changes the necessary to  make dist-sync use ideal CH?

Reading futther, it seems like Infinispan config must have:
<hash numOwners="2" rehashEnabled="false" rehashRpcTimeout="60000" class="org.radargun.cachewrappers.EvenSpreadingConsistentHash"/>

I think RadarGun could do these modifications without forcing users to have a separate dist-sync-*.xml file.

Also, if useEvenHashing="true", can't the code handling the config take the benchmarks numOfThreads and assign it to threadsPerNode?

Also, you're saying that WebSessionBenchmark.numOfThreads and keysPerThread must be the same but in the wiki they're not. Also, as per my last comment, if there's a relationship between keysPerThread and the benchmark config, couldn't the RadarGun figure that out?

Thanks for writing this up!

Cheers,

On Feb 2, 2011, at 12:50 PM, Mircea Markus wrote:

> Hi,
>
> I've just added: https://sourceforge.net/apps/trac/radargun/wiki/MeasuringDistribution
>
> Cheers,
> Mircea
> _______________________________________________
> infinispan-dev mailing list
> [hidden email]
> https://lists.jboss.org/mailman/listinfo/infinispan-dev

--
Galder ZamarreƱo
Sr. Software Engineer
Infinispan, JBoss Cache


_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev
Reply | Threaded
Open this post in threaded view
|

Re: [infinispan-dev] benchmarking ideal distribution

Mircea Markus

On 2 Feb 2011, at 14:47, Galder ZamarreƱo wrote:

Some comments, in:

name="idealdistribution/dist-sync-ideal-distribution.xml" useEvenHashing="true

Does dist-sync-ideal-distribution.xml differ in any way to dist-sync.xml? The reason I ask is cos reading the wiki, it seems like useEvenHashing "makes Infinispan plugin to use the ideal consistent hashing implementation rather than the configured one".
Good point. I've removed useEvenHashing and leave the hash to be configured in ISPN. 
So, if dist-sync-ideal-distribution.xml and dist-sync.xml are different, in which way are they different and way? Seems like RadarGun could already make the changes the necessary to  make dist-sync use ideal CH?
Now they would be different in the consistent hashing class used. 

Reading futther, it seems like Infinispan config must have:
<hash numOwners="2" rehashEnabled="false" rehashRpcTimeout="60000" class="org.radargun.cachewrappers.EvenSpreadingConsistentHash"/>

I think RadarGun could do these modifications without forcing users to have a separate dist-sync-*.xml file.
+1. Just that right now Radargun relies on the name of the file in order to include it in the report, so two different files should be configured.

Also, if useEvenHashing="true", can't the code handling the config take the benchmarks numOfThreads and assign it to threadsPerNode? 

Also, you're saying that WebSessionBenchmark.numOfThreads and keysPerThread must be the same but in the wiki they're not. Also, as per my last comment, if there's a relationship between keysPerThread and the benchmark config, couldn't the RadarGun figure that out?
Indeed! https://sourceforge.net/apps/trac/radargun/ticket/20

Thanks for writing this up!
Thanks again for the excellent feedback!


_______________________________________________
infinispan-dev mailing list
[hidden email]
https://lists.jboss.org/mailman/listinfo/infinispan-dev