Advantages and Disadvantages of Using MongoDB vs CouchDB vs Cassandra vs Redis

Advantages:

Mongo

Lightening quick.
Auto sharding on it's means (August)
Replication is incredibly straightforward.
You can perform made queries, will produce on the fly indexes with one command.

Couch

Very persistent, no equipment failure, system crash can do something to your knowledge.
Everything is hypertext transfer protocol. you'll serve a CouchApp from info itself victimization views.
Replication is straightforward, and may be uni or bi-directional.
Conflict resolution may be a breeze. For distributed systems, this is often a requirement have.
Uses telephone unit, you'll run CouchDB even on your small phone!
Guys at Meebo created Couch-Lounge, a sharding tool.
Everything on disk. terribly sturdy.

Redis

Lightening quick.
Replication potential.
Cool datastructures obtainable among realm of key-value.

Disadvantages:

Mongo

Baca Juga

Very unreliable. No single server sturdiness. If one thing crashes whereas it's change 'table-contents' - you loose all you knowledge. Repair takes loads of your time, however sometimes lands up in 50-90% knowledge loss if you are not lucky. therefore solely thanks to be absolutely secure is to possess two replicas in several datacentres.
Indexes take up loads of RAM. they're B-tree indexes and if you have got several, you'll run out of system resources extremely quick.

Couch

Couch maintains a distinct document for each update you create. this is often done to save lots of all revisions of a document. though this is often a positive feature, this fills up your magnetic disk quick. you'll run asynchronous compaction to delete all recent revisions, however this usually takes hours and is system intensive. however this is often a trade-off for redundancy.

Redis

Memory solely.
You can have slaves, however it is not distributed however. though they will add it in future.

Just needed to summarize some crucial commonalities and variations that were unseen of alternative answers:

1. Cassandra and Couch ar eventually consistent, which means that a shopper may perform Associate in Nursing update, receive Associate in Nursing acknowledgment, one more shopper acting a scan won't essentially see that update. Your applications might or might not be ready to trot out this sort of behavior within the knowledge store. several applications, like commercialism concert tickets, wouldn't work with this model. Others, like searching carts, might maybe work ok, presumptuous that their users will clear up occasional inconsistencies.

2. MongoDB is inconsistent by default. The documentation claims "strong consistency," however the default implementation considers Associate in Nursing operation "complete" as presently because it is queued within the send buffer of a shopper, even before it's been seen by any node. there's a giant discrepancy between the discussion in Mongolian monetary unit blogs and therefore the default implementation within the code. you have got to require further measures to make sure that Associate in Nursing update has propagated to all or any replicas.

3. Redis is not like the others in this it's a single-core in-memory knowledge store. It provides an expensive interface to common knowledge structures, however it's extremely designed to support those knowledge structures for one native method or a collection of processes on identical native host. it's not a distributed store.

4. HyperDex may be a distributed store that provides a powerful consistency guarantee: each GET is certain to see the results of each antecedently acknowledged place.