Personal tools
You are here: Home Virgo Data Analysis VDAS vdas_20110629
Log in

Forgot your username or password?
Document Actions


by Marie Anne Bizouard last modified 2011-07-08 23:14

VDAS call

  • Announcements
  • Current issues:
    • Data transfer (Livio)
    • RDS production (Benoit)
    • Feedback from ASPERA meeting (Gergely)
    • CNAF pilot jobs: summary of last week meeting (MAB)
    • CMT upgrade at Cascina? (Franco)
    • Define new lab CMTSITE (MAB, Franco): CMT packages installation anywahere should not require any requirements customization

  • Specific topics:
    • CMT/autoconf interoperability:
    • Adv Detector software developments:
      • definition of a computing model
      •  workflow
      • Data Analysis white paper: computing & software


Participants: Livio, Franco, Stefano, Alberto, Benoit, Loic, Cristiano, Gergely, MAB

1. Data transfer:
Livio: VA3 missing files at CNAF (raw/proc) transfered done. Start to transfer missing files at Lyon. VSR4 transfer OK (pb of last week end at Lyon fixed thanks to Jean Yves Nief). Has prepared a script that compares CNAF and Lyon rawdata and prepare the transfer of missing files between the 2 places.
MAB: comparising CNAF and Lyon is fine, but most of the diffs are between Cascina and CNAF and Cascina and Lyon. For VA3 gaps most of them were identical in Lyon and Cascina.
Livio: correct and understood. Level of monitoring for VA3 was low. For science run, monitoring scripts are on.
MAB: need to worry of .ffl files construction. New tape/disk system at CNAF might prevent to build dynamically these files

--> action item for MAB and Livio

2. RDS production
Benoit: already produced 10 days. Available to LDR transfer. will send an email to VDAS and Scott K. Request to check everything is OK. Will also produce .ffl file at Cascina.
MAB: production planning for the rest of the run?
Benoit: a production mid July and then the rest at the end of the run.

3. Feedback from 2nd ASPERA Computing meeting
Gergely: computing models for future presentations. LHC computing models more defined than LIGO or Virgo ones:
   - production computing
   - user based data mining
Our "production" part is limited to h(t) and RDS production. The rest of the searches are done through user based developed pipelines and that's going to be the case for the future. In the LHC experiments the fraction of "production" jobs is larger.
Presentations of LIGO (Patrick Brady): try to define a unified framework for the future but this still shows how non uniform are our computing models.
Gergely presentation:  try to show that we do not have a unique solution through the different user cases (CW, GPUs, CBC, ...)
MAB: any discussions with CC representatives?
Not so many discussions with CNAF and CCIN2P3. Ask only how we define our needs ....
Discussions about what experiments are doing. Bruce Allen mentioned Hannover plan to purchase new hardwrae: GPUs ...
Require for Virgo to setup a more defined computing model?
MAB: we have an heterogeneous model. The actual danger is to loose our independence. The use of Virgo resources is decreasing over the years. One of the reasons is the interconnection between condor and LSC(/virgo) pipelines. Except very few Virgo pipelines (CW mainly) all new Virgo developments are done within condor and hence run only on LSC clusters
Gergely: because LSC clusters provide a user friendly environment compared to Virgo CCs
MAB: correct. Need to correct the situation. Long term projects. Need to provide a user friendly environment. If we do not do anything right now, that will be too late in 2 or 3 years.

Gergely: pycbc development (Duncan B. + UWM + Hannover). Goal: make lal and lalapps independant from local disk organization on LSC clusters.

4. Pilot jobs at CNAF: can be launched from a LSC cluster or directly from CNAF. MAB needs to contact group chairs to find users ...
otherwise this development will be given up.

5. Db compilation pb /CMT upgrade:
agreement to
- upgrade CMT after VSR4
- cleanup requirements files
- keep the backward compatibility for old CMT on RIO
- install the new CVS plugin for upgraded CMT

In the meantime, the Db requirements file will be modified to let Benoit release Fd for low latency data transfer test on LIGO sites

6. Next meeting in 2 weeks. Amaldi conflict?