Even for a twitter recommendation engine, that's just a lot of moving parts.

I hope your development is driven by accurate profiling and benchmarks. It's much better to have a dirty, but very surgical fix in a specific bottle-neck, than to have your application "scaled" with too many layers of individually "optimized" components, that bog each other to a crawl when combined.

There are a lot of moving parts, but these are separate sub-systems with well defined boundaries of interaction. Each sub-system is implemented using technologies selected for specific characteristics.

Mono for long-running and complex backend processes.

nginx, PHP, MySQL, memcache - for presentation - quick!

node.js provides an event bus between PHP and Mono

MongoDB is taking over from storing large blobs of JSON in MySQL.

Makes sense.

I was on the opposite side of things, when I did we last. I had language integrated into the web server, and I wrote my own caching primitives. All optimization went into the request handler, and optimizing a single function is easier than debugging a whole stack.

A decent solution, but it had no massive inputs, so I slept better at night knowing it wasn't being tested.

No. You, sir, are not normal. I'd love to know what sort of evil hackery you're cooking with that setup, though.

I'm cooking up a Twitter Recommendation Engine for users, hashtags, and links etc.

For example: http://tagwalk.com/tag/MySQL

Played with it a bit. I like it :D

Thanks!

I am in the same boat, though not with your exact setup. It's easy to acrue alot of individual components, especially if you are a tinkerer, when developing a non-trivial application.

rails, sinatra, mysql, resque, custom email server (looking for a name...), memcached.

How the heck do you mix in mono?!?

nginx ftw