使用瓶子进行路由的性能优势？

Question

I started developing a web application based on the bottle web framework a couple of years ago. I chose Bottle at that time because it was the easiest solution to get up and running fast, and I was just working on building a prototype of an idea I had. Now, I have several thousand lines of code, and I'm looking to move into production-level solution.

At first I thought that I should move to a full stack framework such as django or web2py . As I was moving that direction, I started just using various parts of other frameworks as I needed them. For example, I implemented web2py's data access layer (DAL) so that I could run my application on google-app-engine , and now I'm looking to into using web2py's scheduler to manage jobs. Then, I began to use cherrypy as a production-level webserver. I tried the rocket server, but I was getting more errors with it, so I preferred cherrypy over rocket.

I started looking into rewrite my code to fully use web2py's full stack solution; however, the amount of time to rewrite my routing functions to fully move into web2py seems to be significant, and I really wasn't happy with the rocket server.

I really like the simplicity of Bottle, the way that Bottle uses decorator functions to map the routes to the functions, and also the philosophy of extensibility.

I want to know if there are any specific advantages in terms of performance in using Bottle to do routing, versus any of the full stack frameworks.

I appreciate anyone's advice on this!

Answer 1

According to this benchmark and others, Bottle performs significantly faster than some of its peers, which is worth taking into account when comparing web frameworks' performance:

1. wheezy.web........52,245 req/sec or  19 μs/req (48x)
2. Falcon............30,195 req/sec or  33 μs/req (28x)
3. Bottle............11,977 req/sec or  83 μs/req (11x)
4. webpy..............6,950 req/sec or 144 μs/req (6x)
5. Werkzeug...........6,837 req/sec or 146 μs/req (6x)
6. pyramid............4,035 req/sec or 248 μs/req (4x)
7. Flask..............3,300 req/sec or 303 μs/req (3x)
8. Pecan..............2,881 req/sec or 347 μs/req (3x)
9. django.............1,882 req/sec or 531 μs/req (2x)
10. CherryPy..........1,090 req/sec or 917 μs/req (1x)

But bear in mind that your web framework may not be your bottleneck .

Answer 2

From the creator of web2py, Massimo Di Pierro:

If you have simple app with lots of models, bottle+gluino may be faster than web2py because models are executed only once and not at every request.

Reference:

groups.google.com/forum/#!topic/web2py/4gB9mVPKmho