===================================
GridPP Technical Meeting 2014-11-07
===================================

Present: Daniela Bauer, David Colling, Simon Fayer, Janusz Martyniak,
Alex Richards, Andrew Lahiff, Andrew McNab, Sam Skipsey, David Crooks,
Ewan MacMahon, Gareth Roy, Kashif Mohammad, Peter Love, Chris Walker
plus a couple of anonymous people


1. ATLAS - Peter Love
---------------------

- Changes to Operations

  Frank Berghaus moving to CERN Fellowship

  Moving from Puppet for image contextualization to cloud-init

- Waiting on Icehouse upgrade from Datacentred at Lancs

- Issue with Imperial site name change, which still needs to be
  resolved 



2. CMS - Andrew Lahiff
----------------------

- See slides

- Plan to extend to other sites, including target share testing with
  Vcycle

- Question over extra overhead with VMs because the CVMFS cache is
  started from scratch each time

  Plan to re-run the test with retention of the cache

- More Meyrin vs. Wigner efficiency testing

  Persistent finding that Wigner jobs running ~4-5% lower efficiency

  However, looked at another way, throughput at Wigner was better
  (events per HS06 second)

- Peter Love:

  At Lancs, seen massive loads on Squid caused by Vac

  This should be investigated further


3. LHCb - Andrew McNab
----------------------

- Using new pilot in VMs, and plan to use it for batch as well

- Vac/Vcycle support unit in GGUS

- Plan to use RabbitMQ for monitoring, ElasticSearch and Pica (?)


4. Storage
----------

- Upcoming presentation from Alistair Dewhurst in storage meeting on Ceph as an
  object store via xrootd, from ATLAS perspective


5. Networking
-------------

- Nothing to report


6. AOB
------

- Timing of meetings

  Doodle poll

- Format

  Expanded focus, not just clouds