The previous link had some questions about performance and equivalence for in-memory use for redis vs. MongoDB. I can tell you that we tried to make MongoDB do

The previous link had some questions about performance and equivalence for in-memory use for redis vs. MongoDB. I can tell you that we tried to make MongoDB do high write updates, and a 24 server cluster (8 shards with 2 replicas per shard) was unable to keep up due to the massively slow response times. I replaced it with a single server running a twemproxy cluster of 4 redis instances and that single server was able to handle more than 10x the load. So, there are many cases where using redis makes a lot more sense, even if it makes it complicated to maintain a persistence model.