I actually like the redis-objects gem, which maps redis structures to ruby objects.

https://github.com/nateware/redis-objects

Would you please explain why you prefer it to rb-redis?

redis-rb maps redis commands into the ruby namespace.

redis-objects is built on top of redis-rb and creates objects that are backed by redis, with operations like push, pop, etc. cleanly mapped.

In other words, it's a more ruby-like API.

I recommend reading the github README for redis-objects, it goes into a lot of depth.

The critical distinction of redis-objects is that it does not abstract away the atomic redis commands, which are the source of redis' power. It contrasts with ORMs where the goal is primarily convenience; with redis-objects the goal is power.

I've never used redis-objects but it does look interesting. I'll have a play around with it tonight and update the post.

Thanks for the heads up.

Edit: updated.

Nice intro piece. Especially liked the breakdown of data structures in redis, comparing them to Ruby.

I really loved the "TL;DR: Redis is fucking awesome."

I started using Redis for a project with Node, and I loved the node_redis library because it's non-blocking :-)

I'm working on bringing em-hiredis up to spec. If you want a similarly non-blocking ruby alternative. https://github.com/koudelka/em-hiredis

Redis is absolutely beautiful for async networking; the protocol is extremely straightforward. It has completely solved my problem of persisting data from EventMachine programs in Ruby, too.

A couple hundred lines of ObjC got a generic Redis client and a special-purpose blpop client working for my Cocoa apps (I keep libevent in a thread there).

I would love to see performance info for a site running redis on top of FusionIO hardware.

Redis' usual mode of operation is to keep the dataset in memory, making the FusionIO SSD hardware kind of redundant. With the new diskstore back-end, though, and really fast solid state drives, I bet that Redis could be really fast with a data set much bigger than memory.