I just finished what I would call a small django project and pretty soon it’s going live. It’s only 6 models but a fairly complex view layer and a lot of records saving and retrieving.
Of course, forgetting the obvious huge amount of bugs that will, probably, fill my inbox to the top, what would it be the next step towards a website with best performance. What could be tweaked?
I’m using jmeter a lot recently and feel confident that I have a good baseline for future performance comparisons, but the thing is: I’m not sure what is the best start, since I’m a greedy bastard that wants to work the least possible and gather the best results.
For instance, should I try an approach towards infrastructure, like a distributed database, or should I go with the code itself and in that case, is there something that specifically results in better performance? In your experience, whats pays off more?
As a personal contribution: I sometimes have the impression that some operations, when done through django’s signals, are faster then the usual view way. But hey, I’m biased. I freaking loooove signals. 🙂
Personal anecdotes like mine, are welcome as a way to stimulate some research, but some fact based opinions are much more appreciated. 🙂
Thanks very much.
here is what we did…
In general, don’t try to optimize performance up front. First, collect usage/performance stats, then pick off the most rewarding changes (effort vs. benefit) until you get decent results. The goal should be to avoid unnecessary complexity (distributed databases, etc)
Then, if you still aren’t happy, consider these (in order): more RAM (goes a long way), a dedicated database server, load balancing multiple app servers (using perlbal, etc), a dedicated media server, etc…see these for more details (deployment guide, performance tips)
good luck…