[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Filling pool breadth-first with partitionable slots (Beyer, Christoph)



LIGO doesn't actually generate that much data by 2016 standards. Roughly speaking: 2 instruments each producing a voice-quality-compressed podcast. Campus network operations is a lot more concerned by YouTube.

The users, on the other hand, generate lots of data and my cluster private network considerations are dominated by that. I also have ~100 users active at any time and their generated data are not shared (and therefore not cached in the means you describe). Spreading the load out has its benefits even though I'd rather fill each machine 1 at a time.

--
Tom Downes
Senior Scientist and Data CenterÂManager
Center for Gravitation, Cosmology and Astrophysics
University of Wisconsin-Milwaukee
414.229.2678

On Mon, Oct 31, 2016 at 11:25 AM, Michael Pelletier <Michael.V.Pelletier@xxxxxxxxxxxx> wrote:

Since youâre dealing with the entire universe at the CGSA, I suspect your data sets are much larger than mine. *chuckle* Most of our inputs have no trouble fitting in the systemsâ buffer cache and so N-1 jobs on N cores read them from there instead of the network (also I have âatimeâ turned off on the NFS mountpoints), and a lot of our systems have 10Gb since the Brocade switch per-port cost and the incremental cost of an 10Gb HP FlexLOM hasnât been prohibitively high.

Â

I gather in 8.6 network bandwidth will be a manageable resource â thatâll be nice.

Â

ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ -Michael Pelletier.

Â

From: HTCondor-users [mailto:htcondor-users-bounces@xxxxxxxxxxx] On Behalf Of Tom Downes
Sent: Monday, October 31, 2016 12:16 PM
To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Subject: Re: [HTCondor-users] Filling pool breadth-first with partitionable slots (Beyer, Christoph)

Â

I like depth-first owing as it is an anti-fragmentation tool, but breadth-first is, in my experience, better for input file transfers in the real-world because my storage can easily swamps the 1GbE network interface on an N-core machine run N jobs.


--

Tom Downes
Senior Scientist and Data CenterÂManager
Center for Gravitation, Cosmology and Astrophysics
University of Wisconsin-Milwaukee
414.229.2678


_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@cs.wisc.edu with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/