After reading the paper by Dave Jaffe and Michael Yuan on Seam performance, I was wondering if the authors had more data on number of concurrent users.
For example, if you had 25,000 concurrent users, what would be the performance figures. (I am assuming 25,000 concurrent testing threads would mean about 150,000 -200,000 real users... which would be traffic on a normal consumer site.. Does anybody has concurrent user figures on popular consumer web sites?)
Also the graphs show response times > 1 sec. What is the upper bound of the response times in the graphs. For example, during stress testing 30 percent of the response times are > 1 sec. How much more than 1 sec. are the response times.