Thank you Thadeus. This seems a good approach. Massimo
On Jul 21, 4:33 pm, Thadeus Burgess <thade...@thadeusb.com> wrote: > I try. > > I have several copies of web2py core that I have modified in attempt to fix > these issues, attempting to narrow down the cause, to no avail yet. I want > to do two more tests. The first one, make flask use the web2py DAL so we can > compare core to core. Make sure web2py has all of the fancy optimizations in > place. > > The second thing I want to do is develop a very minimalistic WSGI > application that is not based on any framework. I want to execute the > controllers in two different ways, I want to test a import based controller, > and then an execfile based controller. Very simple, no templates, no > database. This small test will verify if it is an issue that can be fixed > (such as too much overhead processing the request), or if it is an issue > that cannot be fixed (ie: a design flaw in using execfile). > > Once the question of execfile vs import on performance is answered will > depend on what should be tested next to figure out what is causing these > issues. > > -- > Thadeus > > On Wed, Jul 21, 2010 at 9:59 AM, mdipierro <mdipie...@cs.depaul.edu> wrote: > > Thank you Thadeus for your help with these tests. > > > I did not see a open_session(request) in your Flask code. > > > Where does Flask save sessions? filesystem or client? If they are > > saved on the client that is closer to session.forget() in web2py since > > filling the filesystem with one question file for every http request > > when running ab may cause problems. > > > Massimo > > > On Jul 21, 9:44 am, Thadeus Burgess <thade...@thadeusb.com> wrote: > > > I will make said changes, and add my Flask-DAL extension to the flask > > app, > > > so at least the database layer will be the exact same. > > > > Flask handles sessions too. I shouldn't disable web2py sessions while > > > letting flask use sessions, that would be an unfair test too. > > > > The good news is, that even with migrate=True the DAL still outperforms > > > sqlalchemy (yay). So using sqlalchemy would result in a slightly slower > > > flask. > > > > -- > > > Thadeus > > > > On Wed, Jul 21, 2010 at 9:02 AM, mdipierro <mdipie...@cs.depaul.edu> > > wrote: > > > > I agree with your assessment. The problem you, Iceberg, Rahul, and > > > > MikeEllis are having with Rocket appears distinct from the problem > > > > Thadeus is having uwsgi+cherokee. > > > > > In my previous email I tried to suggest changes in Thadeus code to > > > > isolate the cause of his problem. > > > > > Going back to your problem. I am still suspicious that this is a cron > > > > issue. Can you try change cron=True into cron=False in web2py.py? > > > > > Massimo > > > > > On Jul 21, 8:41 am, Kuba Kucharski <kuba.kuchar...@gmail.com> wrote: > > > > > > You said: "I moved to apache/WSGI from Rocket. For me the problem > > > > > > disappears." therefore your problem is not the same that some other > > > > > > problems experienced with apache/WSGI. > > > > > > look at Iceberg, Rahul, MikeEllis in this thread > > > > > > >You also said you cache.ram all > > > > > > requests. Can you try remove the caching? Any improvement? > > > > > > I think there are two or more problems. One is Thadeus having. The > > > > > second was described by me. They can be the same but they seem not to > > > > > be. As I investigated in this thread people talking about "my kind of > > > > > the problem" report using rocket/cherrypy. Removing caching gives no > > > > > improvement.
