Hi, My debug logs show that imports take quite a long time. But Google App Engine spends several seconds even before my first log entry. The whole execution is also slowed down (even without datastore operations).
During the last 10 days, I've been testing various app designs work during the latency periods (I still see them almost every day). 1. The first thing I did was moved the home page request (acccountable for ~30% of app regquests) from Django to WebApp with Django template (different request handler files in app.yaml) The WebApp w/Django requests took roughly 3 times as little to execute as normal Django requests during the high latency periods. This still meant 50-100 seconds per loading request. 2. I moved another frequent request (also ~30% of all requests) to a new request handler in a separate file. This time it was WebApp without Django templates. This request just calculates an MD5 hash and sets a cookie. Very simple. During the high latency periods it took about 25 seconds to load. Faster than requests with Django dependencies, but still too slow. 3. Converted the WebApp request handler to a pure CGI handler with no WebApp dependencies. The only imports are *os*, *sys*, *datetime *and * hashlib* During the high latency periods, this simple request takes 15-20 seconds to execute. How can this be explained? Now I have 4 different requests handlers that use *1*. Django *2*. WebApp w/Django, *3*. WebApp without Django, *4*. CGI and can clearly see how Django contributes to slow execution times. Here are the average executions times (*ms* - *pending_ms*) for such types of requests during the recent high latency period. Django : *132*-*300* seconds WebApp w/Django : *22*-*50* seconds WebApp without Django: *10* seconds (cron task, one dataset put()) Pure CGI : *17*-*18* seconds Normally all requests take less than *1* second (less than *0.1* second for WebApp and CGI) and their loading requests take less than *4* seconds even for Django. So, Django is a significant overhead during the high latency periods. But even simple pure CGI requests load unreasonably sloooow (*15*-*20* seconds for requests that only set a few headers and write about 20 bytes of data). I don't like the idea of not using Django. The project will be hard to maintain. At some point I may consider moving to a different framework, say WebApp2 + Jinja2 or mako (by the way, any recommendations?), but this won't solve the whole high latency issue. Now I think about migrating to HRD. Not sure it will help though. P.S. Anyway, moving 60% of requests out of Django, helped reduce the number of instances that App Engine loaded during the high latency periods. Now the billing history pages show less instance/hours than I had before those high latency periods :) -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/hoo6Iao5IxQJ. To post to this group, send email to google-appengine@googlegroups.com. To unsubscribe from this group, send email to google-appengine+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
