Benchmark code for Access to ODBC

************************************************** ************************************************** ************************************************** *
This email and any files transmitted with it are intended solely for the use of the individual or entity to whom they are addressed.
************************************************** ************************************************** ************************************************** *

Here's a good one for last thing Friday :-)

We are evaluating some ODBC drivers for our Solaris server, to work with
SAS/Access to ODBC (SAS 9.1). The databases we are connecting to are
Oracle on Solaris and SQL Server (needless to say, on a Windows server).
As part of this I need to devise some suitable benchmark tests. Clearly
these will be SAS programs in some form or other, including straight SAS
data and proc steps working on ODBC libraries, PROC SQL code or
pass-through SQL.

Can anyone suggest suitable types of things to test? I don't think the
end users are sophisticated SAS programmers - their code normally
consists of subsetting data steps, sorts, merges, means and tabulates
and I don't think they know about SQL - but I want to test beyond what
they do. It's just general ideas that I'm looking for so no need to go
to too much effort.

Many thanks in advance for this.

----------------------------------------
Nigel Pain
Scottish Executive
Analytical Services Team
Victoria Quay
EDINBURGH
EH6 6QQ
UK
Tel +44 131 244 7237
Mob. +44 7795 618362
Mailto:nigel.pain@scotland.gsi.gov.uk
Website: http:\\www.scotland.gov.uk <http://www.scotland.gov.uk/>


The original of this email was scanned for viruses by the Government Secure Intranet (GSi) virus scanning service supplied exclusively by Energis in partnership with MessageLabs.

On leaving the GSi this email was certified virus-free

Nigel.Pain@SCOTLAND.GSI.GOV.UK wrote:
> We are evaluating some ODBC drivers for our Solaris server, to work
with
> SAS/Access to ODBC (SAS 9.1). The databases we are connecting to are
> Oracle on Solaris and SQL Server (needless to say, on a Windows
server).
> As part of this I need to devise some suitable benchmark tests.
Clearly
> these will be SAS programs in some form or other, including straight
SAS
> data and proc steps working on ODBC libraries, PROC SQL code or
> pass-through SQL.
>
> Can anyone suggest suitable types of things to test? I don't think the
> end users are sophisticated SAS programmers - their code normally
> consists of subsetting data steps, sorts, merges, means and tabulates
> and I don't think they know about SQL - but I want to test beyond what
> they do. It's just general ideas that I'm looking for so no need to go
> to too much effort.

I don't have a good test suite handy.

But you do. I would recommend that you really start with the code
your end users are normally submitting. Just fix the blatant errors.
Then increase by an order of magnitude the size of the data sets and
subsets and outputs involved. (The real data will get that big in no
time.) If you create a test suite which actually mimics the sorts of
tasks done there, then you are more likely to end up with decisions
which actually relate to your group's workflow. I mean, who needs a
decision that only helps if your company process flow happens to look
like an InfoWorld testbed?

David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician