The maximum number of users, 1000 in your example, is really not important. This is a simple, classical, queueing system problem. The engineering criteria is the average number of milliseconds between requests, and the service time of the average request.
Are your estimates of 200kb using bytes or bits? Most networking quotes are in kilobits/second or megabits/second, so if you are talking about 200 k-Bytes, that is about 1.6 million bits. if you have a T1 line from your telco, which is 1.544 mega-bits/second, you can not provide a typical page under your 1 second goal.
Calculation of mean queue length is easy once you know the request inter-arrival time and the mean service time.