I've experimented with increasing the batch size in my application and found, that performance suffered significantly with batch sizes > 30.
So I've tried to find where the bottlenecks are. I've place $start := miliseconds In fbx_ settings.a4d of the root circuit, which should be the very beginning of the request and several writebr (Milliseconds - $start) throughout the code. The last one in lay_main.a4d after </html>, which should be the end of the request. With batch size set to 100 I get around 700 ms for the last statement, which is not blazing fast but acceptable, as there are quite some database queries involved in generating the content. Anyhow it takes much longer until the page appears in the browser (Safari/Firefox). The benchmark tool in Safari tells me, that latency is about 800ms, which correspondents roughly with the the 700ms from my benchmark, but loading time is 15s, which is definitely too long for a good user experience. The test was done on a local setup, sever and client connected via GB ethernet, so internet speed is not a limiting factor. Reported page size is 350 kb, which should be transferred in a fraction of a second over the LAN. I'm wondering what is going on to take 14s for sending the page. Components in involved in the test setup: 4D V13 server on Win 7 64 bit A4D latest NTK latest 4D Client on Mountain Lion Any tips to find and probably remove the bottleneck are appreciated. Peter _______________________________________________ Active4D-dev mailing list [email protected] http://list.aparajitaworld.com/listinfo/active4d-dev Archives: http://active4d-nabble.aparajitaworld.com/
