<ul>
	<li>Local: Andrea Sciabà, Markus Schulz, Michel Jouvin, Domenico Giordano, Evangelia Santorinaiou</li>
	<li>Remote: David Lange, Oxana Smirnova, Renaud Vernet, Martin Gasthuber, Concezio Bozzi, Johannes Elmsheuser, Pepe Flix</li>
</ul>

<p>Markus: we should review the pre-packaged workloads to check they are still current.</p>

<p>Domenico: Eva joined as technical student, will work on containerisation. There is a tool being developed by the CVMFS team to make snapshots. We could put in the container also the input data. The containers can be put in the Gitlab docker registry. The CVMFS team also worked on thin containers for the cases with access to CVMFS.</p>

<p>Markus: we also need reference analysis jobs because 80% of the CPU is used by analysis.</p>

<p>Johannes: we never tried to capture analysis jobs as they are so diverse. We could however use a couple of jobs to cover the extremes of the spectrum, for example a SUSY analysis and a cross-section measurement, one with lots of sparse reads and the other with continuous reads.</p>

<p>Renaud: I contacted the Tier-1 people for the infrastructure costs and I received four answers so far: three to ask "by when is it needed" and a full answer from NIKHEF. David Groep provided a document with lots of details, but only current expenses, not the past ones. I'll send a reminder in a week.</p>

<p>Oxana: I forwared the request to Mathias, but no news so far.</p>

<p>Markus: more and more work is done in the DOMA subgroups: how can we keep up to&nbsp;date? For example, Ilija Vukotic made a study about the impact of site caches similar to our own. We should make sure this information is propagated to us.</p>

<p>Martin: I attend also the DOMA QoS meetins, their goals are aligned to ours.</p>

<p>ACTION: add a task about activities in other groups.</p>

<p>David: in 2-3 weeks we should have a new CMS reference workload.</p>

<p>Concezio: I will review the LHCb workload.</p>

<p>Markus: the GEANT team will provide reference Geant4 (and later on GeantV) jobs. The reason is that they suspect that Geant4 jobs embedded in experiment frameworks may be "polluted" by experiment code and thus not very representative of Geant4 behaviour.</p>

<p>Andrea: a new HEPiX working group on technology watch has started: it is expected to be relevant also for us.</p>

<p>ACTION: Markus will present his estimates on storage costs at the next meeting.</p>