Keywords: DataAccWG, TechAssessWG
Hello,
See below, instead of through attachment, you can fetch the
spreadsheet from:
http://www.slac.stanford.edu/~becla/tmp/lsst_diskIO_estimates_v14.xls
Jacek
-------- Original Message --------
Subject: CPU estimates for query load
Date: Thu, 20 Jul 2006 15:36:12 -0700
From: Jacek Becla <[EMAIL PROTECTED]>
Organization: Stanford Linear Accelerator Center
To: Jeffrey Kantor <[EMAIL PROTECTED]>
CC: Tim Axelrod <[EMAIL PROTECTED]>, Ray Plante <[EMAIL PROTECTED]>,
Don Dossa <[EMAIL PROTECTED]>
Jeff
I'm attaching the spreadsheet with CPU estimates.
The number of CPU needed is 5,329 (on old AMD Opteron 1.8 MHz).
As we concluded on the phone, the numbers seems quite
acceptable =)
See "cpuPerFetch" page for information what query
I used, also see low-vol and high-vol pages columns W-AA
for per-query-type breakdown.
It will also give you a rough idea how much parallelization
we will need to finish a single query in requested time
(although the numbers will be much lower for faster
cpus in 2013)
cpuLoad page is a high-level summary.
The number of disks did not change.
It would probably make sense to rename the file now
from diskIO to queryLoad estimates.
Jacek
_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data
