[Halld-offline] Fwd: Scrubbing computing requirements

[Mark M. Ito]

FYI, from Chip Watson.


-------- Original Message --------
Subject: 	Scrubbing computing requirements
Date: 	Thu, 28 Aug 2014 09:46:56 -0400
From: 	Chip Watson <watson at jlab.org>
To: 	Graham Heyes <heyes at jlab.org>, Mark Ito <marki at jlab.org>, David 
Lawrence <davidl at jlab.org>
CC: 	Sandy Philpott <sandy.philpott at jlab.org>



All,

The reviewed tape requirements for Physics for FY14 indicated a need to 
write 3 PB to tape in FY14.  So far in FY 14, Physics has written around 
1.3 PB, so wrong by a factor of two.

I would like someone to scrub this data.  Is this factor of 2 due to a 
uniform high estimation going on so that you never use more than you ask 
for?  If so, then I would request that you give me two numbers by 
quarter for the next 3 fiscal years: (1) most likely usage, (2) high 
estimate (likelihood 25% or less).

Similarly, we have not yet provisioned disk capacity at the level 
presented in the review, and no one is screaming.  This is also off by a 
factor of 2-3.  Again, is this coming from cautious overestimating, or 
an unanticipated delay in ramping up simulation?

I am now making decisions that are influenced by these projections, and 
so I need better insight as to how to treat your numbers.  (E.g., should 
I treat all numbs as "highest possible need" and simply divide all 
numbers by 2 to obtain "most likely" ?)  GlueX is the driver for the 
next 18 months, so I would especially like GlueX to scrub their disk and 
tape data.

Since FY15 is going to be tight (according to almost everyone), I would 
really like to understand what you NEED vs what you want "just in case", 
since over provisioning by a factor of 2 means less money for Physics to 
do something else. Please also clarify if all of the numbers can be 
scaled with number of weeks of beam on target (and tell me what your 
assumptions are for that).  I.e. try to separate the uncertainty of your 
estimation process from the uncertainty in number of weeks of running.

Please get this data back (or a first cut at it) to me within two weeks 
as it is needed for planning for FY15.  Please pass this request along 
to whomever needs to be involved in this process.

thanks,
Chip



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20140828/06447789/attachment-0002.html>