Scalability issues in RHQ 4.9 - A summary

Heiko W.Rupp hrupp at redhat.com
Sat Mar 8 08:39:42 UTC 2014


Elias,

Am 08.03.2014 um 02:59 schrieb Elias Ross <genman at noderunner.net>:

> I've encountered the following scalability issues:

Thanks for all your investigations and also this summary list.

I know that you are following master.

We are trying to get 4.10 out the door (it is overdue), so I would suggest that
we all start with 4.10 as the new baseline and continue from there, especially
as we have incorporated a larger number of changes already that should
increase performance and scalability.
Having said this, I know that we need to improve further.

We are also thinking of moving the storage node code into its own
repository (and then pulling them into RHQ like other 3rd party deps),
so that should make it easier to contribute there.

And talking of StorageNodes: *I* think that we should (optionally?)
move the computation of the aggregates, baselines etc from the RHQ
server into the StorageNode, so that if you have SNs distributed over
multiple CPUs, that the processing load is also distributed to where the
data is.
Wrt Baselines: *I* also think our current algorithm is sub-optimal and will
need to be replaced by something better in the future (perhaps even pluggable)

Thanks again
   Heiko



More information about the rhq-devel mailing list