Benefits of Completing a TCA

Benefits of Completing a TCA

Now that we have laid the groundwork for a clearer understanding of the concept of concurrent users, the next question is: What level of server resources does your site require? We find the TCA methodology provides a scientific approach to addressing this question.

Because TCA focuses on server resource costs relative to user operations, it can be used whether your Web application is based on the .NET architecture or Windows DNA. The methodology is virtually the same in both cases. The only difference we have found are some slight variations in Performance Monitor counters for Windows DNA versus .NET Web applications. Table 9-1 outlines the performance counters and performance objects in each case. Note that this is not a complete list of counters you will need to monitor. It is only a list of those that have changed locations from Windows DNA to .NET.

Table 9-1. TCA Performance Counters for Windows DNA and .Net

Performance Counter	Performance Object under Windows DNA	Performance Object under .NET
ASP Requests/Sec	Active Server Pages	ASP.NET Applications
ASP Execution Time	Active Server Pages	ASP.NET
ASP Wait Time	Active Server Pages	ASP.NET
Requests Queued	Active Server Pages	ASP.NET

TCA is an excellent approach for focusing development efforts on reducing hardware resource costs for your site s most expensive user operations, thereby increasing application scalability without increasing your hardware infrastructure.

Using TCA to reduce relatively high site operation costs will not only increase the scalability of your site but will also go a long way toward decreasing your site s total cost of ownership. This can be achieved by introducing a reliable method for modeling your site hardware capacity needs as your site traffic evolves and grows. A thorough TCA allows you to maximize site hardware resources already in place, to project when you will need to add more hardware, and to predict how much extra capacity, in terms of concurrent users, that extra hardware will adequately serve. In this manner, hardware costs can be more accurately associated with user population levels. The following real world example describes how a TCA approach assisted in increasing the scalability for Microsoft s shop.microsoft.com site.

Real World Example Shop.Microsoft.com

In the spring of 1999, the ACE Team completed a TCA for Microsoft s eCommerce site http://shop.microsoft.com version 2.5. We found that the heavily trafficked Default homepage and Browse operations had relatively high CPU resource costs. This indicates that users requesting the homepage and product overview pages exacted a higher cost in terms of server resources when compared to other operations on the site. Identifying these high cost operations through a TCA allowed the developers to focus on reducing the site s CPU utilization. Subsequent to completing a TCA on the http://shop.microsoft.com version 4.5 code, we found that the shopper operational cost decreased by 25 percent between the versions, increasing the site s concurrent shopper capacity by the same percentage. This indicated that the shop site could adequately serve 25 percent more shoppers using the same hardware configuration and the version 4.5 code. The majority of the decrease in total cost could be attributed to cost reductions in the heavily trafficked Default homepage and Browse operations that were identified in our earlier version 2.5 TCA. These operations realized a lower cost in the version 4.5 code because they could achieve much higher ASP request per second throughput rates without significantly increasing the CPU utilization required to sustain those throughput rates. Additionally, our original version 2.5 TCA model allowed the developers to perform what if scenarios against subsequent versions by quickly calculating the impact of coding decisions on overall site capacity. The lesson learned in this TCA exercise was that if you understand your Web application s user traffic and subsequent hardware resource consumption, optimizing high resource cost application code will buy you more capacity without increasing your hardware budget.