DOMA / QoS Meeting

Wednesday 23 Jan 2019, 15:00 → 16:00 Europe/Zurich

Present:

Andrea Sciaba
Chris(tophe) Haen
Edward ("Eddie") Karavakis
Gene Oleynik
Laurent Duflot
Mario Lassnig
Martin Barisits
Martin Gasthuber
Mihai Patrascoiu
Njardi
Oliver Keeble
Paul Millar

Project ARCHIVER

Martin Gasthuber introduced project ARCHIVER, an EU funded project,
coordinated by CERN.

Can we follow up here on QoS for archival?
 
This is a PCP project, so need to describe what they want in order to produce a tender.
 
The project has just started and has a 3 year duration.  There is "almost" a website -- it is just awaiting the official release.

The project will have three phases: design, prototype, pilot.  It is currently in its design phase.

April is the first call within the project for the design phase.
 
Martin will circulate more details

Technical/software engagement

Paul placed some suggestions on the twiki page.

Comments:  Move Mihai to EOS; Eddie is name for FTS.

Do we know who to contact from  CTA / StoRM?

Oliver to communicate with CTA.

We can invite them, saying there's no direct need for engagement, but they are more than welcome.

Site survey

Paul:  DESY has completed the survey, with some suggestions on how it may be improved.
 
Oliver: will have the answers for survey for CERN by the end of next week.

There was then a discussion on survey tools

  A simple twiki page has been used in the past, but there are some authentication problems at CERN
 
  Mario suggested using just googledocs?
      Mario to investigate

Paul also suggested investigating some online tools.

VO engagement

Chris mentioned he is happy to contribute but not on a continual basis.

The promotion of cheaper, less reliable storage would need work at VO level, requiring automated data recovery at global level.

He felt that the RAID question hasn't been properly answered: what are
the real tradeoffs here?  capacity, reliability AND performance.

He also noted that automated recovery is being discussed in the DOMA Access WG.
 
Chris also felt that the smaller experiments will not want lots of diversity on their infrastructure, as this can get into a combinatorial explosion.  Therefore, there should be a small number of classes -> this is precisely the classification question.

Oliver - suggestion on single meetings with VOs to come after the white paper is completed.
 
Chris/LHCb liked both the idea of a white paper and the interview idea.

Mario reported that Atlas making progress on the carousel improving sync between workload and data management.