Chapter 9: Estimating IIS Tier Capacity with Transaction Cost Analysis

Chapter 9

Estimating IIS Tier Capacity with Transaction Cost Analysis

Jonathan Swift said, Necessity is the mother of invention; but when you re performing capacity planning, Mark Twain s words are more applicable: Necessity is the mother of taking chances. Why? Because capacity planning is the art of reducing the probability that a Web application s performance will degrade as a result of increasing or changing traffic patterns, while simultaneously avoiding unnecessary hardware purchases. Microsoft s Transaction Cost Analysis (TCA) is a science-based methodology to help you estimate Web application hardware needs. But you can never predict with 100-percent certainty what a user population actually ends up doing with your Web application. Prepare for surprises.

Capacity planning is not about absolutes; it is about intelligently preparing for probabilities. For example, if your company is planning to advertise their Internet address during the Superbowl, a TCA approach will help model hardware resource needs for the increased user traffic you d expect from such a large marketing event, reducing the need to run performance tests for every possible traffic scenario. The TCA approach to capacity planning is nothing new. It has been around since the advent of client-server technology. However, the TCA methodology presented in this chapter is specifically adapted for Microsoft Web applications and services. It was developed in 1998/1999 by a team of Microsoft engineers including Hilal Al-Hilali, Morgan Oslake, David Guimbellot, Perry Clarke, and David Howell. There were a few business reasons for developing a TCA approach to capacity planning back then which are still relevant in today s .NET environment. Firstly, TCA is a scientific approach to estimating server hardware requirements. Secondly, as mentioned above, it enables Web application owners to model site capacity what if scenarios, greatly reducing the number of performance test iterations needed to define required hardware resources. For example, what happens to your Web application s performance if an additional 10,000 users are browsing in response to an e-mail marketing campaign? TCA enables you to answer that question without the need to set up and run additional performance tests. Lastly, TCA helps focus developers on optimizing the most costly code in terms of hardware resource consumption, to increase application scalability on existing hardware. Our team has successfully used the TCA methodology for Microsoft Web sites, including http://shop.microsoft.com.

TCA defines hardware resource costs relative to user scenarios, and these costs are used to model required growth in hardware capacity. The ultimate end product of TCA is a set of user operation server resource costs that can be used to predict your Web application s maximum concurrent user levels. Since the TCA defines maximum concurrent user levels and the term concurrent user is often misunderstood, it merits a brief discussion before we proceed in describing the TCA methodology.