I wrote a new C++ framework (lithium) last december ranked #4 in this benchmark and while this benchmark does a great job comparing raw performances of many fra

I wrote a new C++ framework (lithium) last december ranked #4 in this benchmark and while this benchmark does a great job comparing raw performances of many frameworks on a set of test, here are my thoughts: 1/ Best Frameworks are all using Non blocking network IO (based on linux epoll or equivalents) 2/ Very few companies need such performances. Don't look for those number if you have less than 10M visitor/months 3/ Tuning the number of sql connections can do x5 on performances and all framework did not tune this parameter. 4/ Even if you a low ranked framework, it may not be the bottleneck of your stack so changing to a top performer may not change your thoughput overall.