How to scale CherryPy?

I did a web search on how to scale CherryPy server and didn't find much information. I was wondering if there is a guideline on this subject. We are planning to run two CherryPy instances for a consumer facing application. The backend caching and the static files cachine are already handled, we just need to handle a large number of simple GET requests.

1. how do we scale the front-end?
2. By default cherrypy server.thread_pool is 10. When I increase it to 50 or 100 and run my load testing against it and it seems to freeze the server. Most resources I found are using some number between 30-50.
3. Other techniques for scaling 开发者_Go百科to thousands of users at the same time?

Thanks!

---

Here are two sites you should take a look at...

http://yougov.github.com/pycon/slides/

http://www.readmespot.com/question/f/100994/cherrypy--load-balancing--and-remote-failover