Print

Print


A "short" comment:

The Xrootd team has been advertising virtual mass storage system  (MSS) for
a long time. The idea is to use the Xrootd's MSS interface to fetch data,
not just from tapes, but can also be from other sites dynamically. In the
case of ATLAS, the fetching process will use ATLAS DDM tools.

I had a long talk with Doug Benjamin about this idea last month and it looks
like he is interested to have T3s be able to fetch data dynamically in this
way, and he later e-mailed back and asked us about the status and
development needs. To my knowledge, the existing xrootd and xrootdfs have
interface to do this without further development. All we needs are scripts
using DDM tools and following the interface requirement. We do, however,
need more discussion with Doug about expected behaviors and operational
mode.

regards,
Wei Yang  |  [log in to unmask]  |  650-926-3338(O)



> From: "Kim, Peter Chong-Ho" <[log in to unmask]>
> Date: Thu, 7 Jan 2010 16:13:44 -0800
> To: atlas-sccs-planning-l <[log in to unmask]>
> Subject: Minutes of  ATLAS/SCCS Planning Meeting: Jan. 6, 2010
> 
> 
>  ATLAS SCCS Planning, January 6, 2010
>   ----------------------------------------------
> 
>     9:00 AM, SCCS Conf Rm A,
>     to call in +1 510 665 5437, press 1, 3935#
> 
>     Att:  Charlie, Richard, peter, Tofigh, Randy, Booker, Wei
> 
> 
>  FY2009 Hardware Install
>  -----------------------
>     CPUs (part of ~190 units) will be installed at Forsythe.
>     After consulting with DOE, Teri is working on an MOU.
>     Target date for installation is end of January.
>     Dedicated fibre connection to SLAC has been ordered.
> 
>     All 10 Thors are now in production.
>     The first 5 that went into service need to be moved to another
>     rack; Wei is coordinating outages with SCCS and ATLAS/PANDA.
> 
>     We need to start planning on FY2010 purchases with an expectation
>     of more storage needs. Thors with 2TB disks are not yet
>     available, but will be soon. Other alternate storage options
>     were mentioned in previous meetings.
>     The decision deadline for timely ordering/delivery/install is
>     end of June, 2010.
> 
>  Tier-2 Operations
>  -----------------
>     Running smoothly. ATLAS took great advantage of fair-share
>     during the Winter break as other contingents were not active.
>  
>  Pilot Study of HPSS in ATLAS
>  ----------------------------
>     Richard mentioned the pilot HPSS storage test for local users,
>     starting with ~100 TB. (For 1 TB tapes, that's 100 units.)
>     To be really useful, one would need a system of order 1 PB
>     (1000 tapes).
>     ATLAS storage management software has to interface with SLAC
>     HPSS system which was mainly developed by BABAR. Andy and Wilko
>     are working on xrootd issues; Wei on other software.
> 
>     Some interest was shown by D.Benjamin at Duke on testing HPSS,
>     though the main focus of HPSS is mainly on storage capacity
>     rather than active running of jobs via staging.
>          
>  NFS server for releases, T2 operation
>  -------------------------------------
>     The current server is an old/slow T3 that needs to be replaced.
>     Wei will look into the option of solid state disks and very large
>     memory system for fast access.
>  
> 
>  Cooling Tower Upgrade
>  ---------------------
>     No news on the status of rebidding; the target completion date
>     after the initial delay was set as April 2010.
>     Bldg. 50 is ready for connecting to temporary cooling units
>     which could provide enough cooling in winter.
> 
>  Cavern background
>  -----------------
>     Charlie requested dedicated resources at SLAC for generating
>     cavern background events. The scope of CPU usage and storage
>     requirement needs to be fleshed out, but Tatsumi and J.Black
>     will start with general queues for submitting jobs.
>     For the outputs to be transported to a central location (CERN)
>     and accessed by many users, they should be registered to central
>     catalogue.  
> 
>  Future Meeting Time
>  -------------------
>    To avoid time conflicts with other ATLAS meetings, it's been
>    suggested to move the ATLAS-SCCS meeting to another time
>    and/or day of the week. Andy Hass will take a poll to find
>    a new time slot.
> 
> ----------------------------------------------------------------
>