The code in models is executed at every http request. The longer the
models, the more slow down. Yet the slow down you get is larger than I
would have expected. I did some tests about this long ago and I found
this issue to be negligible. Let us know what you find and I will take
a look too. Meanwhile, what os are you suing?
On Jan 7, 9:44 pm, "James M." <sipajah...@gmail.com> wrote:
> I just did some simple load tests on a web app I am working on. I was
> using apache bench (ab -n 1000 -c 100) and getting results that were
> much slower than expected. The Requests/Second were ~ 26. For
> comparison I did the same test against the welcome app and I was
> getting ~ 59 Requests/Second (with no optimizations).
> I started testing various scenarios to determine what was causing the
> big discrepancy. I tried all the suggested tips for improving
> performance (migrate=False, compile app, cache, session.forget() etc),
> with very small improvements...
>
> I was curious what kind of overhead adding tables to the welcome app
> would have. To get a good starting point I decided to optimize the
> page I was testing in the web app as follows:
>
> #default.py
> @cache(request.env.path_info, time_expire=60, cache_model=cache.ram)
> def index():
> session.forget(response)
> return response.render(dict())
>
> After this change I was getting 93 Requests/Second
>
> I added the following table:
>
> db.define_table('table1',
> Field('col1','string'),
> Field('col2','string'),
> Field('col3','string'),
> Field('col4','string'),
> Field('col5','string'),
> Field('col6','string'),
> Field('col7','string'),
> Field('col8','string'),
> Field('col9','string'),
> Field('col10','string')
> )
>
> and tested again resulting in 87 Requests/Second
>
> I repeated adding tables (table2, table3 etc)
> After adding 5 tables the test was down to 67 Requests/Second
> After adding 10 tables the test was down to 52 Requests/Second
> After adding 30 tables the test was down to 27 Requests/Second
>
> The full test results are logged
> here:https://docs.google.com/spreadsheet/ccc?key=0ApWaSSISueScdG1DSDl4NVNk...
>
> I am starting to look at the DAL code to see if I notice anything that
> can be improved, haven't noticed anything yet. I am wondering if
> anyone has any thoughts on what would be causing this.
