Atlas-connect-l mailing list

[Rob Gardner <rwg AT hep.uchicago.edu>]

Harinder,

On Jan 16, 2014, at 11:21 AM, Dr. Harinder Singh Bawa <harinder.singh.bawa AT gmail.com> wrote:

Hi Rob,

I agree: It would be great to have weekly meeting to know the developments in "atlasconnect" and if there is any other urgent needs to address to any issues.

Yes indeed.  We are still struggling to catch up here after our network upgrades and our new CPU deployment.  I would like to say we could meet tomorrow, but at this late moment perhaps its better if just plan on NEXT Thursday, 9 am PST for sure.

1) Dial Toll-Free Number: 866-740-1260
2) Enter 7-digit access code: 4766859 followed by “#”

Yongsheng maybe you can send us a reminder. 

1. Just for info: One of the main concern for me was our Tier-3 bandwidth for the data transfer during the condor jobs. At this moment its 1GB but scheduled to increase to 10GB in summer probably. But I wanted to check by submitting the condor jobs in our cluster which reads the input file from Tier-2 (using continuous bandwidth) and I used FAX system. I am happy to see that its working at this moment. Although the jobs are not very high in number but it works according the job slots available on our CSUFresno ATLAS cluster. 

Attaching the network plot during of the job. 

A very interesting result.  This is just the sort of use-case we’d like to pursue.  Besides the network bandwidth used, a interesting figure would be the additional time required to process a dataset via FAX, versus the case where you’ve stored all the files locally.   If its taking only 50% longer, then thats a good result.  On the other hand, if it takes x10 or even x100 longer, then well it might pay to copy the dataset.   Also, note the use of TTreeCache is needed.

Do you know which FAX site(s) were used in your analysis?

2. If you remember, we had discussion before about the environment difference between clusters. I observed if i submit the job through "atlasconnect" the default os is slc6(may be my area) and CSUFresno Atlas cluster, we have 5 nodes on slc5 and 2 on slc6.  This was leading to my jobs failed in some nodes.But now after I get to know the real reason, I modified my submission script to check first the os and setup Athena/Root accordingly. 

Wonderful news. 

The above tasks were on priority for me (as presented in Arizona workshop) since I can use "atlasconnect" in doing the analysis using ROOT/athena now and I see FresnoState cluster working great at this moment with whatever resources we have.

Absolutely - this has been my impression too - I’ve been able to get lots of jobs through the Fresno cluster, a resource I would otherwise not had any access to.  Thats pretty cool, in my book.

I will continuously doing some more rigorously testing involving analysis and data…

Very good!!

BTW - I have to give a FAX review on Feb 3.  Using FAX in this fashion demonstrates a particular capability not otherwise available. If you have the chance to summarize your studies, perhaps in a single slide or two, that would be useful to include in the talk.   Note - it would be very interesting to try your analysis with your input data sitting at SLAC and compare the result to the case when it is at MWT2 or some other place.  Perhaps you’ve done this.

Best,

- Rob

Thanks & Regards

Harinder

On Tue, Jan 7, 2014 at 1:35 PM, Rob Gardner <rwg AT hep.uchicago.edu> wrote:

Hi Yongsheng,

Yes indeed.  This week is tough for us as we’re still working on the network upgrade where the atlas-connect-l services live.   Lets maybe plan on one next week.  I’ll send around a doodle poll for the specific time.

Cheers,

- Rob

On Jan 7, 2014, at 1:46 PM, Yongsheng Gao <yogao AT csufresno.edu> wrote:

Hi Rob and others,

Happy New Year!

Harinder is doing tests as discussed in the Arizona workshop. Perhaps we could resume our weekly ATLAS Connect meeting soon. As new semester is starting with different working and teaching schedule, perhaps it is a good idea to pool the mailing list again about meeting schedule. Thanks.

Best,

Yongsheng

---

Rob Gardner • Skype rwg773 • 312-804-0859 • University of Chicago

_______________________________________________
Atlas-connect-l mailing list
Atlas-connect-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/atlas-connect-l

--

Dr. Harinder Singh Bawa

                                          
[web][facebook][youtube][twitter]

<ganglia-network-condor.png>

---

Rob Gardner • @rwg • Skype rwg773 • 312-804-0859 • University of Chicago