Accurate benchmarking

I'm currently experimenting with table fragmentation and PDQ on our HP K200 running HPUX10.01 and DSA 7.31.UC1. I've been getting some promising returns at the SQL level with PDQPRIORITY set to 1 and OPTCOMPIND either 0 or 1. (Yes, at the SQL level I've gotten some impressive gains using sequential scans instead of index scans. Whether or not I'll implement that type of change remains to be seen.) Currently, I've been running a shell script that cycles through a series of SQL statements. I'm in an isolated instance, so I can drop the engine after each SQL; that way I can test disk accesses. I'm also running timex for each dbaccess statement. While I know that this setup will give me the time until I access the first 'screen' of data, I'm curious: Is there a more accurate way to benchmark this test? I've thought about trying to pipe the output to /dev/null, but I haven't been able to combine that along with the command-line syntax. That would allow the SQL to run to completion. Any other ideas out there? Or is this accurate enough to show gains / losses in PDQ and table fragmentation? Thanks in advance, John Carlson Informix DBA WH Smith, Inc. BTW: I'm testing the SQL now. When I've narrowed down the field I'll be testing full application. -------------------==== Posted via Deja News ====----------------------- http://www.dejanews.com/ Search, Read, Post to Usenet