Tobin Bradley has posted about the speed difference between ArcIMS and Mapserver. “Although I can’t rerun their test to validate the results, I can anecdotally support them - I have certainly found MapServer to perform significantly faster than ArcIMS, and that experience has been echoed to me by other parties. The study found MapServer to be about 30% faster than ArcIMS, which sounds about right.” Wow, 30% faster! That gets your attention, but when looking at the posted results in his post (he says he found them on a listserv, but Google wasn’t kind to me today) you see on average that UMN MapServer is half a second quicker than ArcIMS in the performed test. Not exactly anything to be amazed about and probably not measurable by most people and certainly not significant.
I won’t argue with anyone who says UMN MapServer is fast because I’ve seen it, but in our real world applications we haven’t noticed any difference between our applications running on ArcIMS vs those running on MapServer. Benchmarking something like that is a complete waste of our time, plus I don’t own a stopwatch. The bigger issue with the speed of both GIS server applications is how and where your data is installed. It is easy to get caught up on these speed claims with server software and they are fun for posting to Digg, but in the real world there are way too many variables to worry about 1/2 a second waiting for a map to be served.
ArcIMS does has some issues, but I don’t consider speed to be one of them.