On 30/01/13 11:02, emyr.james wrote:
> Dear all,
>
> The Sussex saga is nearing a successful conclusion. We seem to have most
> things working properly and reasonably reliably now, and atlas testing
> is well under way.
> My immediate plans now are to get some local monitoring and metric
> measuring in place, and to try to do some performance tuning.
>
> I'm starting off by going through Steve Lloyds test pages
> (http://pprc.qmul.ac.uk/~lloyd/gridpp/ukgrid.html)
>
> If you look at the network tests
> (http://pprc.qmul.ac.uk/~lloyd/gridpp/nettest.html), our performance is
> quite poor at the moment.
For the column (that's you copying to other GridPP sites),
http://fasterdata.es.net/host-tuning/linux/ is a good place to start.
For the row, your NAT gateway, or local storage is performing poorly.
> I have contacted our network guys to arrange a
> meeting to look into this. One thing that I've noticed is that there is
> no data for Susx/Susx (i.e internal copying). What do I need to get in
> place for this to start being picked up by this test ?
>
> We are missing from the following pages...
>
> http://pprc.qmul.ac.uk/~lloyd/gridpp/metrics.html
> http://pprc.qmul.ac.uk/~lloyd/gridpp/hs06.html
> http://pprc.qmul.ac.uk/~lloyd/gridpp/uktest.html
> http://pprc.qmul.ac.uk/~lloyd/gridpp/hammercloud.html
I've asked Steve Lloyd about those. Essentially they _should_ all
appear automatically once there are results.
If Steve remembers correctly, the UKtests require the latest release
that 50% of sites have ( and you don't yet).
>
> ...and seem to perform poorly on on....
>
> http://pprc.qmul.ac.uk/~lloyd/gridpp/releases.html
> http://pprc.qmul.ac.uk/~lloyd/gridpp/benchmarks.html
>
At least one of the install jobs failed:
https://atlas-install.roma1.infn.it/atlas_install/jobs.php?relfk=366557
That might be you, might be the atlas install system.
> If anyone can suggest anything to improve our exposure/performance on
> these tests that would be marvellous.
Come along to the Thursday Atlas UK meeting.
Chris
|