===================================
GridPP Technical Meeting 2014-11-21
===================================

Present: Andrew McNab, David Crooks, Gareth Roy, Jeremy Coles, Peter
Love, Wahid Bhimji, David Colling, Daniela Bauer, Simon Fayer, Duncan
Rand


1. ATLAS
--------

- Nothing to report


2. CMS
------

- David Colling:

  More large scale tests on T0 and Wigner

  More work on HLT

  Agreed that HLT will be used in inter-fill periods

  CMS Open Data release

- Andrew McNab:

  Has VMs running on Vac VMs at Manchester, running CMS MC jobs (based
  on Andrew Lahiff's work)

  Looking at what's needed to make them work on cloud sites as well

  Applied for a Vcycle development tenancy at CERN OpenStack

  Intending to run multi-experiment workloads to test target share
  mechanism

  To create a form of "excess spot market"


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

- More work on pilots inside VMs, on job completion verification

- LHCb Vcycle service at CERN migrated to officially managed LHCb
  vobox at CERN

- In Manchester, moved GridPP cloud Vcycle instance to permanent,
  dedicated machine


4. Sites
--------

- Nothing else to report


5. Storage
----------

- Wahid:

  ATLAS testing deliberate overflowing to increase federation usage

  Duncan gave example of a job running this way that was very
  inefficient

  Some of the strange behaviour may be down to running user analysis
  jobs

- David Colling:

  A lot of optimisation on data placement was carried out by CMS to
  improve this federated data access

  Lazy download used with T0/Wigner


6. DIRAC
--------

- Daniela:

  Setting up COMET VO from scratch

  DIRAC makes incorrect assumptions about the storage path

  DIRAC developers worked on this but there are problems with their
  implementation 

  QMUL users interested in a file catalogue

  Clear need for proper quality-control/testing in the central DIRAC
  project 

  Process for integrating external contributions is lacking

- Andrew McNab:

  LHCb has its own project, to get around these issues

- David Colling:

  Could the LHCbDIRAC testing framework be reused?

- Andrew McNab:

  It's very LHCb-specific

- Daniela:

  Need to move to better hardware etc.

  Proper upgrade procedures needed

- David Colling:

  * Try to push for better practices in DIRAC development

- Andrew McNab:

  Consider a separate distribution in future?


7. Networking
-------------

- Duncan:

  QMUL first spotted this issue with ATLAS FAX transfers


8. HEP Software Foundation
--------------------------

- Jeremy:

  Talk at GDB

  Synthesised the 11 white papers

  Started contacting major communities e.g. MC generators, Geant4,
  ROOT

  Various Google groups created

  Preparations for workshop at SLAC 20-21 January

- Andrew:

  Slots for short talks by projects at SLAC workshop

  We could consider submitting to that?


9. AOB
------

- Jeremy:

  Traceability working group on clouds, led by Ian Collier

  Mailing list created, Doodle poll for first F2F meeting

  pre-GDB on Volunteer Computing

  28th largest site for simulation in ATLAS is volunteer computing

- Andrew:

  Responses needed for Laurence Field document on cloud activity

- David Colling:

  Big Data Infrastructure Workshop in December


10. Actions
-----------

- David Colling:

  Try to push for better practices in DIRAC development