Question

We started with Redis, storing active data, logged in users, etc. We’re using some pubsub too for realtime data passing.

Recently we added Mongo to fit our geo spatial needs, and it seems great for non-active data too.

How should these two work together? It is dumb to use both? Is it dumb to pass chunks of data from mongo to redis when they becomes active?

Our thoughts were that we might store everything in mongo but then pass user data from mongo to redis when a user is active and the data is likely to be accessed. I know Mongo does some cacheing like this on its own, we are new to both of them and just want to know how they should be used together, if at all.

Thanks!!

Answer 1

It is dumb to use both? Is it dumb to pass chunks of data from mongo to redis when they becomes active?

So I feel like there’s actually a legitimate to test and validate this question. Redis is basically an “in-memory” DB, so how much better can you do by giving that RAM to Mongo?

Historically, we’ve used the Memcache/MySQL combo to basically “add RAM” to MySQL and limit the amount of writing it needed to do. We did this simply because it was complicated to shard MySQL.

However, MongoDB provides a sharding mechanism. So you can “add RAM” to a problem (along with “adding disks”) simply by adding more shards.

Thanks to the way memory-mapped files work, MongoDB tends to keep recently used data in memory. So if you’re pulling recent data into Redis, that data is probably also in memory on the MongoDB side, so it’s not clear that you benefit from having it in two places.

Is it dumb …

That’s hard to say without some testing and analysis. MongoDB doesn’t really have the pub/sub mechanism, but it does tend to have fast query times, so it may be appropriate in specific spots.