Distributed Apps & Performance Testing

There can be unique performance issues introduced when running your application across multiple company networks or in different vendor clouds. Some architects decide to address dispersed user populations by putting a copy of the app in their geography. For instance, LoadStorm runs in Virginia, California, Singapore, and Ireland in order to provide testing that has user traffic originating on different continents.

Any complex data center or network environment can blow up your performance testing models if you don't account for them in test planning and execution. You can get the proper steps built into your scenarios such that the relative volume of user types (anonymous, buyers, admins, etc.) are represented. That said, you probably want to test the scenario volume coming from a corresponding geo.

It may be necessary to set up multiple tests because you want to apply load to your European data center from a load generation server in Europe. That is only the case if that reflects reality. I've seen many European sites get 80% of their traffic from the US or an Australian site get most of its visitors from Southeast Asia. It all depends. Look at your logs and figure out how you can make your load as real as possible. It will give you more accurate performance results.

I've had conversations with clients in other parts of the world that were concerned with latency, and it can be an issue. However, in the grand scheme of things, most performance problems are significantly more troublesome INSIDE the application rather than OUTSIDE the application. Latency between continents is usually sub-second. Whereas, database bottlenecks are commonly creating 10x the performance degradation that latency contributes.