Response Time SLAgreenspun.com : LUSENET : Service Level Management : One Thread
What is the best, most accepted way to structure an application end-to-end response time service level?
-- Robert Cicero (email@example.com), November 13, 2002
Robert, It is not clear whether you are asking about how to write an SLA to cover end-to-end response time, how to measure the response time, or what to measure.
If you are interested in measuring the response time, the two most widely accepted approaches are to either instrument the application or instrument the environment. Application instrumentation is very attractive, because it avoids the problems of intrusively instrumenting the environment. However, application instrumentation is not without its pitfalls. It is true that it will absolutely measure the response time of every transaction that is processed. However, the first challenge is to get the application instrumented. If it is a commercial off-the-shelf program, unless you are a massive account for that vendor (e.g., General Motors) you are going to have to live with whatever instrumentation is provided. (That is usually minimal.) If the application is developed internally, it is frequently difficult to get the additional development resources allocated to include instrumentation.
Synthetic transactions are the most popular form of external instrumentation being used today. These are not a “silver bullet.” There are both advantages and disadvantages to using synthetic transactions to measure end-to-end transaction response time. I discussed this in an article that I wrote earlier this year: http://www.nwfusion.com/newsletters/nsm/2002/01236828.html That article should give you some more insight into at least a few of the issues related to synthetic transactions.
If you question is what to measure – in simplest terms, the answer is the user experience. You must take measurements, which together will reflect the total user experience across the full spectrum of transactions and across the breadth of the physical environment covered by the SLA.
I have just begun to scratch the surface on this topic. If you wish to discuss this further, please feel free to contact me directly.
Rick Sturm President Enterprise Management Associates firstname.lastname@example.org
-- Rick Sturm (email@example.com), November 27, 2002.
If you send me email I'll send you a snippet from an SLA that addresses end-to-end response time.
-- Mike Tarrani (firstname.lastname@example.org), December 09, 2002.