Design considerations for hard numbers

The spec gives a few very specific numbers:
1 app server
2 web servers
5 sec agent response time
10 sec web customer response time
200/35 - 600/20 number of users
3 sec TransMaster response time
How do you consider these details in your design?
For example, 5 sec agent response time calls for faster UI than web customer so you make it Swing and make your app EJB-centric.
How do you consider scalability (going from 200 to 600 customers) in your design?

Originally posted by Gennady Shapiro:
The spec gives a few very specific numbers:
1 app server
2 web servers
5 sec agent response time
10 sec web customer response time
200/35 - 600/20 number of users
3 sec TransMaster response time
How do you consider these details in your design?
For example, 5 sec agent response time calls for faster UI than web customer so you make it Swing and make your app EJB-centric.
How do you consider scalability (going from 200 to 600 customers) in your design?

You've to understand that your design should not be specially designed to support 200 or 600 customers. You've to design your application for maximum performance without counting customers. If you design your application correctly, scalability will turn into a "hardware" problem. By the way, Swing UI clients is not a remedy - they are still too slow. That's my 2c.
Alex.