With Memcached and Squid, is there any need for ASP.NET caching?

Question

Hi,

With squid, we can cache webpages. I am not sure if it provides the same number of caching methods as ASP.NET caching (I primarily use ASP.NET), but it's a tool to cache webpages.

Then we have memcached, which can cache database tables. I believe this is correct, and it is like SqlCacheDependency (correct me if I am wrong).

However, is there any situation in a large web application where one would find room to use memcached, squid, AND ASP.NET (or PHP, JSP - application framework-level) caching.

Thanks!

Answer 1

You may find that caching entire pages is too coarsely-grained, and caching database tables doesn't get you enough of a boost, leaving a big need for caching chunks of stuff.

Say, for example, you had an application that showed the name of the logged-in user on every page. Caching entire pages wouldn't really work, so you need to drop down a level and cache somewhere within the app framework.

Answer 2

Then we have memcached, which can cache database tables. I believe this is correct, and it is like SqlCacheDependency (correct me if I am wrong).

Memcached is a distributed hashtable. The main benefits over the built in .NET caching is that your cache is scalable (you can add as many memcached boxen as you want) and synchronized (all your web servers have access to the same stuff, and invalidating or updating data from one web server is instantly propagated to all of them)

Performance-wise, it is worse than the .NET cache (you are looking up objects across servers, as opposed to an in-memory lookup on one machine)

However, is there any situation in a large web application where one would find room to use memcached, squid, AND ASP.NET (or PHP, JSP - application framework-level) caching.

For the reasons above, I can imagine a 2-level cache, using the .NET cache first, then memcached. (e.g. a Get() looks at memcached, stores the result in the .NET cache set to expire in 10 seconds, then uses the .NET cache for all the get calls with the same cache key during the next 10 seconds, rinse, repeat)

This way, you get the performance of the in-memory cache lookup without the network IO cost of a pure memcached solution, with the synchronization and scalability benefits of memcached.