Hi Chad, I had done my first experiments on my development workstation. Since then, different setups of our application have been benched on a replication of our production platform.
This platform is sized to support 15.000 concurrent users using different applications. Our target for the application using Jackrabbit is 2.000 concurrent users. We obtained the following : With davex + webapp 2.1.0 : < 5s up to 450 users - for a CPU load of 20% Then CPU load jumps to 100% and response times increase dramatically. It seems to be caused by a sudden increase in the number of threads. With embedded Jackrabbit 2.1.0 : < 1s up to 400 users - for a CPU load of 25% Then the server stops responding and the CPU load fall down to few percents. I don't know yet what caused this issue. With embedded Jackrabbit 2.2.7 : We've reached 1000 users, < 3s, for a CPU load of 50% We are about to bench with davex 2.2.7. I'm very interested to integrate your patch and check what are the outcomes with our application. I plan to do that ASAP. Le 21 juil. 2011 à 23:59, ChadDavis a écrit : >> Experiment A (davex + webapp jackrabbit + postgres) : >> Sc-read - 10 concurrent users : 0,5s >> Sc-read - 40 concurrent users : 2s >> Sc-read - 80 concurrent users : 3,4s >> >> Sc-write - 10 concurrent users : 3,7s >> Sc-write - 40 concurrent users : 8,5s >> Sc-write - 80 concurrent users : 17,2s >> > > With the new patches I've submitted I've obtained the following: > > ( davex + webapp + derby ) > > 50 concurrent users: < 400 ms > > 100 concurrent users: < 1000ms > > However, at 100 concurrent users, I'm not seeing long term stability, > but I haven't figured out the cause. I actually suspect it's our app, > not the davex stack.
