performance - Estimating maximum users that an application can support -

i analyzing web application , want predict maximum users application can support. have below numbers out of load test execution 1. response time 2. throughput 3. cpu

i have application use case sla response time - 4 secs cpu - 65%

when execute load test of 10 concurrent users (without think time) particular use case average response time reaches 3.5 seconds , cpu touches 50%. next execute load test of 20 concurrent users , response time reaches 6 seconds , cpu 70% surpassing sla.

the application server configuration 4 core 7 gb ram.

going data suggests web application can support 10 user @ time? there formula or procedure can suggest maximum users application can support.

tia

"concurrent users" not meaningful measurement, unless model "think time" , couple of other things.

think case of people reading books on kindle. average reader turn page every 60 seconds, sending little ping central server. if system can support 10,000 of pings per second, how many "concurrent users" that? 10,000 * 60, or 600,000. imagine people read faster, turning pages every 30 seconds. same system able support half many "concurrent users". imagine game halo online. each user emitting multiple transactions / requests per second. in other words, user behavior matters lot, , can't control it. can model it.

so, application, have make reasonable guess @ "think time" between requests, , add benchmark. start approach reasonable simulation. other things think session time, variability, time of day, etc.

chapter 4 of "mature optimization handbook" discusses lot of these issues: http://carlos.bueno.org/optimization/mature-optimization.pdf