Application Performance Engineering Blog – Shunra Software

Anyone who performed a load test (with HP LoadRunner for example) has experienced this uncomfortable feeling when asked “Can you tell me how this application will perform in production? Can you predict application performance for our remote users?”

The reason these questions are so hard to answer is centered around the fact that even though LoadRunner test beds accurately replay the expected end user load and application behavior, they don’t take the network impact into account (that is, until now). This isn’t a small detail, the network (especially the Wide Area Network) between the user and the data center can  impact application performance dramatically. This impact can’t be underestimated as many transactions will experience a two fold to ten fold increase in response time under WAN conditions compared to the same transaction performance of a local user.

Past attempts to address this inaccuracy problem included anything from deploying remote load generators across the production network to limiting the scope of the load tests to focus on back-end server scalability while omitting the important aspect of end to end performance testing. Each of these approaches has serious downfalls, some of which are discussed in this post.

I am glad to announce that this post will introduce a better solution to address the need to perform accurate load tests that can actually predict end user performance of both local and remote users. Not only is it a better and easier to use solution, it is also a more accurate solution with reported accuracy levels of 92% – 97%. This solution is called VE Desktop for HP Software which was co-developed by Shunra and HP. This product is officially released  with the release of LoadRunner version 9.5 and Performance Center 9.5.

I will leave the promotion of this great tool to our marketing department, I would like to focus on the practical aspects of this solution. Specifically, the next series of posts will present various use cases and include a step by step tutorial for each use case.

Here is my proposed agenda, but feel free to suggest additional topics that you would like me to cover:

1. Adding network conditions to a LoadRunner scenario – answering the age old question “How will this application perform for remote users?”

2. Automating a LoadRunner script to analyze transaction performance under different latency levels (also known as a “network latency performance curve”)

3. Adding analysis tags to a LoadRunner script to automatically generate deep transaction analysis reports for each transaction in the script (including the well anticipated client network server breakdown and the application turns analysis)

4. Building automated test plans for regression testing that cover a potentially unlimited number of network sites and conditions

I am looking forward to your feedback on the suggested agenda as well as your feedback from using this new and exciting solution.

Best,

Amichai