ansaurus

Question

Answer 1

+3 A:

By default, CherryPy's builtin HTTP server will use a thread pool with 10 threads. If you are still using the defaults, you could try increasing this in your config file.

[global]
server.thread_pool = 30

http://www.cherrypy.org/wiki/ServerAPI

diatoid 2010-04-21 18:17:23

@diatoid: thank you very much! Btw I thought that cherrypy was built in order to support the highest amount of req/sec possible.

RadiantHex 2010-04-21 18:59:53

CherryPy tries to set sane defaults, which produce around 1200 req/sec max on my laptop. But then, those benchmark requests don't take 3 seconds each. The reality for your site should be somewhere in the middle; if your real requests take 3 seconds each you're probably doing something wrong ;)

fumanchu 2010-04-22 01:18:38

requests take 3 seconds because they are waiting for information to be collected from elsewhere. While waiting I am not using any reasources what so ever! So why do I have to leave my machine idle while I can serve more requests at the same time?

RadiantHex 2010-04-22 15:12:24

You are using resources; for any webapp, you're keeping the socket open, which uses a file descriptor and an ephemeral port. In CherryPy, each child connection is bound to a thread for its lifetime, so you're also using one of the worker threads (and its 1MB stack size and any heap objects needed to handle each request).So you have a choice: either 1) increase the number of threads, 2) redesign your app to return 202 Accepted and poll for the answer (which would free up the socket, too), or 3) use an async webserver (and then fight latency issues instead of throughput ones).

fumanchu 2010-04-26 15:48:22

ansaurus

tags:

views:

answers:

CherryPy and concurrency

related questions