Hi Masahiro,

here is the information I collected so far from Roberto and Kerstin. 
I have not received feedback from Sheila. Hence, I propose to scale 
the disk space used by the current Run1-Run4 analysis to Run1-5.

Cheers,
Heiko


>>> 1) How much working disk space does your AWG have available at
>>> SLAC and/or at your AWGs Tier-A site? How much disk space is
>>> that per active analysis?

Roberto made a quick sum at the end of August, adding up the space 
taken in
/nfs/farm/babar/AWG12/ISL/
/nfs/farm/babar/AWG36/
/nfs/farm/babar/AWG38/
/nfs/farm/babar/AWGsemilep01/

and the result was 782 GB for VubRecoil.

>>> 2) What fraction of the available disk space is your AWG typically
>>> using? If you are usually close to 100%, how often do you have to
>>> clean up, to get running again?

VubRecoil:
In recent times, the fraction has always been close to 100%. Personally,
I (Roberto) had to clean up a couple of times in the last year, and in a 
couple of cases Chukwudi and Michael had their jobs crashing because of 
disk space issues and had to find some other space/clean up.

Unfolding:
Currently, about 10GB of AWG disk space is used at SLAC. This is the
amount we can live with to evaluate systematic uncertainties but it
also means cleaning up on regular basis.

>>> 3) If your available working disk space (per inverse fb of data)
>>> at SLAC
>>> and/or your Tier-A site would be reduced, how would your AWG cope
>>> with that? Assume the relative reduction would be 10% or 25%.

Vubrecoil:
That would be a problem, as we are already running low.

Unfolding:
We would cope with it by moving systematic VVF results to LBL.

>>> 4) Which skims are being used by your AWG? Which of these skims
>>> correspond to currently active analyses (or were recently active
>>> e.g. for
>>> ICHEP 2006 results)?

We use the BSemiExcl skim.

>>> 5) If the available storage space for skims (per inverse fb of
>>> data) at
>>> SLAC and/or your Tier-A site would be reduced, how would your AWG
>>> cope with that? Assume the same reductions as for the working disk
>>> space.

Same as the previous answer. We are already running low and need to find
storage space for the whole of Run 5 (Michael's analysis - and I guess
Sheila's too).

>>> 6) If your AWG is using deep-copy skims, what fraction of the skims
>>> are analyzed at Tier-C sites? Could this fraction be increased?

None. Everything is analysed at SLAC. I (Roberto) tried to run at gridka 
some months ago, but the experience was not successful: too many glitches 
and hang-ups, and eventually I gave up.
(Please note that GridKa is not officially holding the BSemiExcl skim.
 The performance of GridKa has been significantly improved.)

>>> 7) What is your experience with running on skims at SLAC / your
>>> Tier-A site? Do you think the CPU power is sufficient? If the
>>> available
>>> CPU power would be reduced, how would your AWG cope with that?

I (Roberto) think that CPU-wise we are fine. Our jobs run smoothly over 
the entire Run 1-4 dataset in a few days.