oh, interesting. I wanted to use redis for log-parsing, but was scared by the prospect of keeping everything in ram. Ended up using mongodb instead, but I'd be interested to hear how you get on with redis.
If you know what stats you want to collect before parsing, the memory usage is very low. The most resource intensive part in this application is loading the filtered logs for all machines into a sorted set (which is necessary to get all the events in chronological order). That key is only temporary though.
That said, Redis 2.x supports virtual memory so it's less of a problem for a longer-term data-store.
