Re: Performance base line

Hi Rick

I know this is less scientific but back in 1996/7 when i was working for a bank in Vienna, Austria and we implemented a finance system running on top of an Oracle database a lot of discussions took place regarding SLA's and performance baselines. This was complicated by the fact that the company i worked for supplied the software, another company built the database for us in the UK and supported it from there, the users connected using Oracle forms over a large pipe from Vienna and a third company in India supported the application source that we supplied and of course the banks admin staff were in the middle of this... This led to interesting discussions of what was acceptable performance. The upshot is that the base line was developed as a set of user run tests that were timed on various application screens that were in the business critical path i.e they affected the time to process customer deals, the batch processes were also timed and data quantities noted. The baseline was in fact simply screen response times and batch window time (allowing for data quantities). These were agreed after initial testing and tuning efforts.

anyway the point i wanted to make is that you can separate the baseline from the details of the database, application, network, etc but the baseline has to be SLA's on what is acceptable response times from the customer / user view point. These were then tested regularly. I have also seen where a similar method screen scraping was used to automate the screen input tests.

You will still need to collect database, network and server stats as discussed by others to understand why an SLA is broken but the baseline in my opinion should be viewed from the business / user view point of view as well.

kind regards

Pete

-- 
Pete Finnigan
email:pete_at_petefinnigan.com
Web site: http://www.petefinnigan.com - Oracle security audit specialists
Book:Oracle security step-by-step Guide - see http://store.sans.org for details.