>> I could change this to be configurable, so that it would only yield after N >> lines of code, which would effectively give more CPU to Active4D and less to >> the database engine. So you would probably want to do this only on the >> Clients > > That sounds like a very useful option (at least in my case).
I'll work on it next week. > By the way, after further testing, the framework code turns out to be only > around 30% of the hit (i.e. 0.3 x 0.4 = 12% overall hit). The rest is either > extra code that I've added or some specific bottleneck. > Also, my customer informs me that this site is actually very fast as rated by > Google, so we don't necessarily have a performance problem A 12% hit is very reasonable for the vast improvement in development speed and robustness that fusebox gives. The question is always, "Is it fast enough?", not "Is it fast?". > Another interesting thing that came up in tests today is the difference in > behaviour between 4D and NTK - > > - with 4D WS under high loading, the socket error rate is about 40% but > response times on the remaining successful requests is under 5 seconds > - with NTK, using Rob's new function to set the socket backlog queue length > to 128, the error rate drops to zero but response times jump to 25-30 seconds Wait, are you saying the web server actually returns a connection error to the user? Wow, that really sucks. > If anybody reading this has tried caching in Apache with dynamic sites I'd be > keen to know how it worked out. I would search on this topic in relation to PHP or Rails, it's the same problem and same solution. Regards, Aparajita _______________________________________________ Active4D-dev mailing list Active4D-dev@aparajitaworld.com http://list.aparajitaworld.com/listinfo/active4d-dev Archives: http://vasudev.aparajitaworld.com/archive/active4d-dev/
