Redis Vs. Memcached

问题:

I am using memcached right now as a LRU cache to cache big data. I've set the max object size to 128 MB (I know this is inefficient and not recommended) and total memcached to 1 GB. But 128 MB is not enough for my purposes so I am planning to move to Redis. A couple questions:

1. memcached is extremely slow - My current memcached setup is taking 3-4 seconds to return just one request. This is extremely slow. I sometimes need to make up to 30 memcached requests to serve one user request. And just doing this takes 90 seconds!! Am I doing something wrong or is memcached actually this slow?
2. Redis would be faster? - I plan to use Redis lists to cache the data. I'll fetch full lists using 0 to -1. I hope Redis be faster because I might as well not use any cache if its going to take 90 seconds!

Thanks!

回答1:

I'd recommend doing a little profiling to see where the bottleneck is. My uninformed guess is that with such large objects, you may be limited by the connection between your app server and memcached and thus you'll see similar results with redis. It could also be that your app is taking a lot of time marshaling and unmarshaling a lot of objects. If it's easy, it might be worth trying a caching scheme where you're just caching the request being sent down to the client (which I'm sure is much less than 128MB).

Another thing to try would be turning on compression. This would give added latency compressing/uncompressing but would reduce network latency if that is indeed the issue.