Date:         Mon, 15 Jun 2009 06:58:31 -0700
Reply-To:     The Architect <thomas.michael.skinner@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         The Architect <thomas.michael.skinner@GMAIL.COM>
Organization: http://groups.google.com
Subject:      Re: How Green is your SAS Code?
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

> While I worked for awhile for Intel  at SAS on sizing and
> optimization, and they supported that activity, one must realize that
> their licensing model relies on CPU power and so they really don't
> have any vested interest in helping folks reduce ther loads other than
> to keep them satisfied from a broad performance perspective.  Many of
> the organizations I have consulted with have expressed a need to
> reduce platform creep which tends to escalate all of their licensing
> costs. So its up to us to do that due dillegence.
>

I should probably clarify my comments per SAS not having a vested
interest in you reducing your SAS loads...
SAS does provide excellent tools for monitoring resources through
their product ITSV (Formerly ITRM), which now includes support for
SAR.  These tools are generally used for monitoring large arrays of
servers used in Banking and other Financial Services enterprises as
well as Mainframes for capacity and performance tuning.  This can be
an invaluable tools in and of itself for identifying system
utilization rates and providing the data you need to manage your
resources effectively.

It is not directly focused on SAS usage however.  It is possible to
monitor SAS specific processes using SAR and some of the macros
provided by SAS will get you started on collecting such data.  As far
as I know, you will have to build your own analytical processes for
linking this performance data back to actual SAS code running. I have
done this using various techniques to convert SAS logs to a format
that can be merged with SAR data.  It typically involves picking up on
SAS job start times and some mathematical conversions of elapsed clock
times in order to do this.  Of course, manageing the data collection
process is the most time consuming task, but can be automated to a
great extent through scripts.  IF you's like to discuss further with
me on how I did this, please feel free to join me at Linkein through
my link below. I'd like to hear about your techniques as well.

Tom Skinner
http://www.linkedin.com/pub/thomas-m-skinner/6/a31/516