I dont think your question has enough data to answer properly. We need to know whats the service time(avg request takes how many seconds) for each request and h

I dont think your question has enough data to answer properly. We need to know whats the service time(avg request takes how many seconds) for each request and how the arrival rate of your requests like (120K per second or minute?) and whats the limit on your request queue (put them on hold till they get server). Finally whats the tolerance level of the final response time (can support 120K users by serialising across 10 servers but that drives up the response time for end users)? Anyways probably you can read some books here at http://www.cs.gmu.edu/faculty/menasce.html http://www.cs.gmu.edu/faculty/menasce.html or if you are in a hurry a quick glimpse at the tactical paper at http://www.cmg.org/measureit/issues/mit04/m_4_7.html http://www.cmg.org/measureit/issues/mit04/m_4_7.html Hope this gets you started if not answer your question thoroughly..