Hi Chris/All
In the interests of dissemination beyond the mail list and also to
preserve the summaries I'm wondering if more sites would consider
contributing to their Tier-2 or the GridPP Operations Blogs? These and
others are linked from here: http://planet.gridpp.ac.uk/. This worked
very well for the Tier-1 updates during STEP09, an example entry being
http://www.gridpp.rl.ac.uk/blog/2009/06/19/step09-flushing-the-buffers/.
Thanks,
Jeremy
-----Original Message-----
From: Testbed Support for GridPP member institutes
[mailto:[log in to unmask]] On Behalf Of Brew, CAJ (Chris)
Sent: 22 July 2009 12:08
To: [log in to unmask]
Subject: Re: Hammercloud review
Hi Graham,
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Graeme Stewart
> Sent: 22 July 2009 10:24
> To: [log in to unmask]
> Subject: Hammercloud review
>
> For your delight and delectation:
>
> ANALY_RALPP: Good. Real user is running jobs which cause an athena
> crash, which is most of the errors. Will change to dccp staging today.
We think we've probably hit our bottleneck until you can enable dccp -
the two cacti plots attached are the connections to our stack with most
of the worker nodes attached and as you can see we're basically maxing
them both out now.
Interestingly the outbound traffic of ~1Gb/s is probably data being sent
from the storage on the Worker Node stack to be retransmitted back from
a GridFTP door on the other stack. The switch to dccp should essentially
double our throughput.
I'm currently buying the parts to up the link to 10Gb/s, it'll be
interesting to run the test again once that's in place.
Yours,
Chris.
--
Scanned by iCritical.
--
Scanned by iCritical.
|