Attending: Dan, John H, Jens (chair+mins), Brian, Matt, Pete, John B, Winnie, Elena, RobC, Steve, Sam, Alessandra


0. OPerational blog posts

   No operational issues reported (by sites), except Pete's which was resolved thanks to JohnH.  Matt also reported having to
   reboot a head node after squillions of connections.

   However, it was reported that Chris B had "interesting" database issues with his dCache upgrade.

1. Summary of your impressions of last week's GDB.

   * EOS federated - storage federation across a very wide area network, seems to perform well so could work for us,
     but of course the proof is in the pudding.
   * Russian work on dCache, setup a la NorduGrid.  Also interesting for its feedback, e.g. security on control interfaces.
     ... perhaps proposed features for dCache 3...
   * Oliver's data management task force, will also have a significant slot at WLCG

2. Input for WLCG workshop - now we know what registration costs and
Oliver says there will be a data workshop:

https://indico.cern.ch/event/609911/sessions/238310/#20170620

   ... so we can discuss what we might contribute and who would contribute it and all that.  Since it's in the UK,
   there's an opportunity to go and participate; and there's a lot of stuff we could contribute to in Oliver's list.

   - caching
	- ATLAS cache problems?
	- CMS: could do work with Oxford (Sam needs to coordinate with ChrisB), and also the xroot level federation
	- Apropos federation, how does a commercial cloud CDN work?  Presumably it redirects to its regional copy.  Maybe worth
	  an experiment.
   - storage accounting (if Rob A and yours truly could finish their stuff)
   - tape (T1?)
   - performance testing of object stores - if there's interest - cf Brian's email
	- on the topic of testing object stores, Alastair asked for a friendly site to help test T1 Echo (maybe just submit a job to a non-RAL T2?)
	- ATLAS QR suggested lower efficiency on Echo: however, no one from Echo was present in this call, so we need to follow up outside.
   - protocols
	- an obvious home/excuse for migrating from SRM, e.g. if SURLs have "srm" in them and files need to be copied out
	  and back in.

   Alessandra reminds us to register soon [and often ;P] so they can fix the numbers.  However the deadline for registration will be
   (or has been?) extended.  Sam and Brian are already registered.

  There is a US Computing White Paper (= CWP), see Alessandra's links in chat, which is being mooted as a 10 year roadmap for WLCG;
  the "evolution" working groups are also working on input for this.  Some experiments have already provided feedback.  It is
  expected that this document will be presented on day 1 of the workshop, and is expected to be completed June.  Maybe it is not
  realistic?  Presumably a 10 y roadmap needs to have both a vision plus a more realisitc way of getting us there.  Compare where
  we were ten years ago and everyone was hyping clouds; clouds are useful but are not the panacea.  And we have in many ways come
  to the end of Moore's law.

3. KeyDocs!  Goodness me they are still r-r-r-rred! many of them.

4. Brian mentioned the object store performance testing. If people are
interested in discussing its use?

   - Maybe Matt?

5. AOB




Daniel Peter Traynor: (17/05/2017 10:01:04)
hi
Jens Jensen: (10:02 AM)
https://indico.egi.eu/indico/event/3249/timetable/#20170511
https://indico.cern.ch/event/609911/sessions/238310/#20170620
> 
Alessandra Forti: (10:31 AM)
http://hepsoftwarefoundation.org/activities/cwp.html
http://hepsoftwarefoundation.org/cwp/cwp-working-groups.html
https://docs.google.com/document/d/1wEsX2BegSBEZe3EXwdNR9VWNPrNEOVVTLhu-NtH-77E/edit
Jens Jensen: (10:33 AM)
https://www.zurich.ibm.com/BigDataAnalytics/program.html