WLCG Archival Storage Group

Europe/Zurich
31/S-027 (CERN)

31/S-027

CERN

10
Show room on map
Description

WLCG Archival Storage Group

  • The Cost and Performance Model WG
    • Possible collaboration on quantifying costs
  • Carousel R&D
    • Overview of progress and results
    • Site experience and feedback
    • Optimising the performance of dCache systems
  • Update on metric reporting
  • Future WG topics

Participants:

Andrea Manzi, Andrea Sciaba, Aresh Vedese, Bo Jayatilaka, David Mason, David Yu, Dimitros Christidis, Dmitry Iitvntsev, Dorin  Lobontu, Doris Ressmann, Enrico Fattibene, Eric Fede, Gene Oleynik, Giuseppe Lo Presti, Jens (RAL), Lauent Duflot, Mario Lassnig, Oliver Keeble, Paul Millar, Ron, Xin Zhao, Tigran, Vladimir Bahyl

 

Andrea Sciaba Presentation:

- all T1s and T0 will be asked to provide a cost report as the one done at IN2P3.

- the cost of regenerating the data from scratch is vastly inferior to the cost of disk redundancy

- cache impact:  the stageout aspect have not been taken into account

- Pierre ( from IN2p3): the presented slides  for IN2p3 are not presenting the accurate results

- another meeting to discuss the cost model for tape carousel will be needed.

 

BNL update on ATLAS carousel:

- why so difference between BNL and TRIUNF?

- a problem could be that BNL use more level of disk buffers. ( more copies)

- limits on the number of staging request can be set also in dCache of course.

- there is a state of SRM in dCache where the requests are queued. (https://its.cern.ch/jira/browse/FTS-1264)

- PIC would like to set staging limits per activity ( dCache can do fair share per user ) (need this into to be available first https://its.cern.ch/jira/browse/FTS-1284)

- would be nice to show how many tape families have  been used in the test.

- why INFN and FZK have the same rate if FZK has 4 times more drives? ( it depends also on the number of tape mounts performed during the test, there is also a limit of max 2000 staging requests)

- Paul Millar said that the focus should be on sites to understand their bottlenecks

- Aresh said that sites should have some targets to be achieved.

- Paul said if it would be nice to have a kind of Hammercloud for tape recalls. so that site admins can request to test possible improvements done their side.

- Aresh reported that it would be nice to have information about the volume accessed to tune the system

 

Tape Metrics Update

- can you report the read/write bytes in 24h? INFN can do it

- can you report the total mounts and unique mounts? INFN can do it

- to be follow up in the mailing list

 

Future of the WG

- Carousel tests are better discussed in the WG

- RAL would like the WG to continue

- Paul Millar thinks the WG is very valuable

- (after the call) Xin Zhao also thinks that the working group is very valuable as it's the only place where tape related issues are discussed

There are minutes attached to this event. Show them.
The agenda of this meeting is empty