[Hps] JLab computing resources projections

[Nathan Baltzell]

Hi Everyone,

It is time again to give an estimate of HPS’s future computing needs at JLab
over the next year.  I checked our 2014 data against Matt's projections from 6
months ago, and they roughly agree (if I assume a 2x increase in event size
from the SVT), even though we ran with various trigger rates.

Below is a projection for data, pretty similar to last time with some improvements.
Any comments?

Last time HPS projected 6 weeks X 300 cores for simulation in the coming year.
Do people think that has changed?   I don’t have any feeling for this.  
I guess filesize for simulation is small compared to data. 

-Nathan



We expect to get 1 week of data in spring 2015.  

That should be ~16 TB of raw data, and I’m going to say another 200 TB of
processing and reprocessing output.  That allows for two full passes of
reconstruction with the 4x output size, and then some skimming, and some
more processing of our 2014 data.

So let’s say 250 TB of new tape for real data for the next 6 months.


Then we probably want some more disk space.  Currently we have 3 TB on
/work and 500 GB on /volatile, for a total of 3.5 TB.  I propose to ask for 12 TB.
You can see what other JLab experiments have here for disk alottments:
http://scicomp.jlab.org/scicomp/


Keep our batch farm cpu request from last time of 500 dedicated cores.