3hqa wrote:When you're running from the cloud, can you tell me the geographic location of the machine?
In most cases, yes, you can tell from which location the actual machines running the tests are actually located. Contact us for details on how this is done, for example, in one of our PerformanceTest projects.
However, it is only fair to point out that the actual location of the machine is not a dominant factor in determining web application performance. The main variation betwen two servers located "here" and "halfway around the world" is the amount of time it takes to start up an HTTP/S transfer. This time consists of such factors as "DNS lookup time", "time to first byte," and similar metrics that you see being used to illustrate geographic performance variations.
However, once a file transfer is started the gross transfer rate is usually determined more by the output capacity of the server machine, and the input capacity of the client machine, than it is by the speed of the internet that connects the two internet. (Of course there are exceptions, but if you are on a relatively high-bandwidth backbone, any slowdowns in the internet are usually a minor factor compare with the limits imposed by the other two factors.)
The way to think of this is that the connection startup and protocol formalities actually turn out to be a very small percentage of the total time of a page transfer -- a percentage that gets smaller as the page size grows.
When we do PerformanceTest experiments -- driving 100's or 1000's of Browser Users (BUs) from cloud computers -- the input/output bandwith of those cloud-based machines is SO large that that doesn't really represent a slowdown factor. It's the speed of the CPU and the input/output situation of the server that imposes the limits.
--The eValid Team
by 3hqa » Fri Mar 19, 2010 5:45 pm