Attending: Elena, John B, John H, Marcus, Matt, Pete, Raul, Sam, Jens, Brian

Apologies: Tom

Speaking of apologies, Jens also apologised for joining late, having been engrossed in only a single-pipe CASTOR problem.


0. Operational blog posts (and yes, Brian and I are allowed to blog again)

   We currently have only two remaining days of the quarter, and no posts since April!

   Case of running out of semaphores on a system - perhaps caused by the large number of drains on the system.
   Kashif seems to have fixed it - can be monitored.

1. Data focused pre-GDB? and the resurrection of the wlcg-data list - cf
   Oliver's mail.  Suggestions?

   Seems to be a good idea - we should send someone from GridPP, presenting work on behalf of all of us?
   - GridPP document on the future of T2s (incl protocol zoo)
   - Testing caching DPM, report from ATLAS (Alastair's long awaited report), Durham cache
   - Accounting and stuff for DPM and CASTOR: interesting work to report?

2. Data einfrastructure update - climate, UKT0

  - JASMIN (www.jasmin.ac.uk) just held their first conference.  Notable that climate have data problems as well and generate solutions
    which are similar to those of GridPP - but using (sometimes) different technology.  Their end users are non-certificate users, though.
  - Doesn't seem to be part of the UKT0 discussion much although they are connected to it.  Lots of cloud work.
    [Which reminds me our very own Andrew Lahiff is working with them on cloud stuff.]

  - Further on the UKT0 front there has been a paper circulated from ATI (Alan Turing Institute) which focuses on the necessity
    of running parallel algorithms (i.e. multicore, not like GridPP) on high end equipment to see how it performs, and discusses the 
    potential for working with emerging computing technologies.  Not much directly on data but obviously data is relevant in the
    high performance data analysis (=HPDA) discussed by the document.

3. Data-related highlights from last week's hepsysman?

   - Most people voted for Marcus' ZFS talk.  There seemed to be a lot of interest and the talks (and the blog posts) are quite
     detailed so there should be enough to get people started.  If there is further interest/need, we can talk about organising
     more ZFS related stuff, e.g. at the next hepsysman?
   - dCache is back...!
   - Lancaster - Matt restored database to check performance, but Vidyo refused to let us hear the results...

4. AOB

   NOB


Marcus Ebert: (29/06/2016 10:22:02)
I think the current length is still fine since it also contains good examples
raul: (10:22 AM)
Is the document available for reading?
Samuel Cadellin Skipsey: (10:24 AM)
Hi Raul: the document will be made available for reading, as soon as I've made some of the changes that Marcus and others have suggested, I want to get it into a slightly more consensus position before opening it up wider
raul: (10:25 AM)
Good. Curious about it. Even more if the test with cache at Brunel goes ahead.
Jens: (10:28 AM)
www.jasmin.ac.uk
Samuel Cadellin Skipsey: (10:30 AM)
(JASMINE's target for "Phase 3" was 15 PB of total storage. Which looks like it was the end of last year... but it's not clear if that was met?)
Matt Doidge: (10:33 AM)
*shakes fist at Vidyo*
I really enjoyed the ZFS talk.
brian davvies: (10:35 AM)
i was taling about form th esites
John Bland: (10:38 AM)
getting the right hardware for ZFS too, our RAID controllers are annoyingly non-JBOD
Marcus Ebert: (10:40 AM)
yes, we unfortunately also have only non-JBOD controllers which is not optimal....
but still works fine :)