Personal tools
You are here: Home Virgo Data Analysis VDAS vdas_20101215
Log in

Forgot your username or password?
Document Actions


by Marie Anne Bizouard last modified 2010-12-15 18:53

VDAS meeting

Pending action items

1. Data transfer status (Livio)

  -  VSR2/VSR3 RDS transfered at CNAF and Lyon

  - S6 H1 L1 data at CNAF (done)

  - S6 H1 L1 data at Lyon (scheduled)

  - ffl file location at CNAF?

2. Data migration Castor --> GEMSS at CNAF (Karen C.)


3. GPU developments (Gergely)

- LSC meeting resume/minutes

- Virgo projects:

4. VirgoDataTransfer / Grid project (A. Colla, C. Palomba)

- data access tests in CCs (Gergely and MAB)

Attendance: Adam, Gegerly, Romain, Alberto, Cristiano, Livio, MAB


Data transfer:

Livio has
- finished H1_LDR_C02_L2 and L1_LDR_C02_L2 transfer to Bologna
- moved folders to LIGO/S6/

Transfer to Lyon will start in the next hours.

MAB: VA1 and VA2 needs also to be transfered from CNAF to Lyon. Low priority task.

Livio informs us that he will change 2 things in the "permanent transfer software"

1/ 50 Hz and trend transfer to Lyon will now be done with his code. Before it was
managed by Daniel Sentenac. Will contact Daniel to coordinate. Recently 50 Hz and
trenddata are transfered to CNAF  like raw and h(t).

2/ change bbftppro --> globus/gridftp in Livio's code. Motivation: bbftppro no more
maintained at CNAF. It will not change any thing about final location of the files
at CNAF.

FFL files generation at CNAF:

present situation:
- ffl files regularly sent at CNAF with Livio's tool: .ffl file automatically generated.
- when files are moved, no ffl is updated.

- Franco C. is storing in /opt/exp_software/virgo/virgoData/ffl/ some .ffl files that
should be used by data analysis.
Problem: no control/monitoring that these files are updated when files move.
         list not complete. For instance VSR3 ffl files are missing

Livio is proposing to setup a little daemon that would scan regularly folders and update
ffl and especially check if files are corrupted.

action item for Livio: .ffl file generation script at CNAF

Data migration at CNAF in GEMSS

first start with CASTOR migration. MAB answered Karen's question.
then cleanup virgo_gpfs4 before migration to GEMSS.

action item for MAB: will send an email of thing to delete on virgo_gpfs4

GPU developments:

Gegerly and Adam gives a summary of last week's meeting. Many developments from independant
groups. Most of them consist in moving the most intensive computation part of the pipeline
on GPUs. For instance chi2 or Hough transform.

Leone has a more ambitious project: full cbc pipeline on GPU. That has the advantage to
avoid the overhead of the GPU-CPU communication.
--> need to know more about this project.

Budapest is working on chi2 implementation. This is very similar to what UWM is doing:
FFT on GPU to compute chi2. + Hough transform.
Gegerly mentions that for the moment the Hough transform is not running faster, but work
in progress

Virgo NIKHEF is working on a polynomial search on GPUs.

Virgo POLGRAV group (Michal Bejger) is trying to work on the Hough transform implementation.

E@H is implementing the F statitic on GPUs.

MAB: this is very important to present regularly the state of the work done by Virgo groups in LV calls (chaired by Drew K.) and other occasions.

action item for MAB: will contact Leone to understand better what he is doing

Pilot jobs

Romain is reporting pbs to run condor flocking jobs from CIT. It worked fine in sep and oct. Stuart Anderson is on the subject.
Other sites are working well. Carsten is setting up Atlas right now.

Romain: work on wscan on demand using condor flocking mechanism. More next time.
Gergely asked if he tested the simultaneously submission from different sites.

Romain will report on that next time.


MAB is still struggling with VO registration.
Gergely has successed to run a grid job accessing data (copying locally). Will post his script on the wiki.
MAB has started a "grid for dummies" wiki page. Encourage GRID job submitters to post script and infos to help new comers.

No progress on VDT development.

Alberto is asking for any experience with gfile library? Gegerly will answer.

Next meeting: mid january