Want to learn one of those NoSQL databases, looking for advice on which one to check out

Question

As someone who first encountered relational databases and SQL at uni in the early 90's, and who has spend a goodly slice of a fifteen year career working with MS SQL Server, I'm probably a textbook example of the people who, according to NoSQL enthusiasts, "have SQL as my hammer and see everything as a nail."

So, to broaden my horizons, I'd like to do some sort of mess-around project using one of these NoSQL databases, as a learning experience.

Question is, though, which one should I mess around with?

• MongoDB seems to be one of the best known and most respected
• Raven DB? It's getting some buzz and Hibernating Rhinos have made stuff I respect.
• CouchDB, since it's part of the Apache project?
• something else?

I don't have a problem to solve, a project to complete, or a client to satisfy, so I'm not asking which might be best suited to any particular purpose. Just recommendations based on quality, potential, and perhaps most importantly, availability of good docs, guides, tutorials, screencasts, etc. Oh, C# / .NET is my field of expertise, so looking for something which fits that well, e.g. has C# examples in its documentation.

Any tips?

Answer 1

There are several variations of NoSQL databases.. The ones you have listed are from the DocumentStore category. Redis is from the key-value part of the NoSQL family. Fundamentally they both follow the principles of NoSQL, but do things a differently.

Structured data may be better suited for the DocumentStore family compared to Key-Value variation(Redis).

Maybe a good way to test, would be to implement the same project using two(or more) different NoSQL solutions.

Oh, my suggestion would be to actually use RavenDB since it's native .Net. No third party libraries that wrap operations. No, which wrapper library should I use. The mailing list, based on what I recall, is really active and in general I think this project is gaining momentum. Constant updates, new features etc etc. Plus, it's Ayende's baby who I think does not ever sleep - so if yo need support at ungodly hours that won't be a problem -:)

Answer 2

I love MongoDB. I even ended up moving my blog to use it for storage.

MongoDB is a designed to be a scalable, high performance database. It is not designed for situations where strict formats need to be applied. It supports atomic inserts which makes it highly suited for real time data aggregates. It has support for standard indexes, sharding, and even commercial support. It models itself after the idea that databases are specializing and you can't create something to please everyone. It uses the JSON document data model to provide cross service support easily and fast. By removing transactional semantics from standard database design you strip most of the "fat" from traditional SQL processes. This comes at a cost however, you can't create foreign keys or constrain the data like you can in SQL. Systems modeled after transactional systems (think banking or corporate records) won't be efficient in Mongo as they are in other systems.

^{Yep, written by me}

I've heard decent things about CouchDB, however not played much with it myself. I haven't heard great things about it, but Apache projects are for the most part decent.

Standard RDMS are not bad at all. If you need transactional logs and SQL queries go for them. The advantages you have is everything is relational, small size, lots of features, giant community support, and there aren't many surprises. They are solid development tools.

Some people think NoSQL means No - SQL. It's supposed to mean Not Only SQL. There are things out there that help greatly when you need to change the table structure on the fly, or where you might not even know what you'll need to store. This flexibility gives you speed and low overhead, but you loose some of the richness relational databases provide. For instance you can't do cross collection queries in MongoDB (SQL JOIN).

There is a tradeoff, and it's not minor by any means. I would, however, encourage you to play with MongoDB. They even have an online shell that you can test out. (Note: It's a JavaScript shell, you can write functions and loops and fun stuff).

Answer 3

There is also OrientDB a open source document-graph dbms with commercial friendly license (Apache 2). Simple API, SQL like language, ACID Transactions and the support for Gremlin graph language. Uses binary and RESTful/HTTP protocols. The SQL has extensions for trees and graphs.

Answer 4

MongoDB seems to be considered one of the most easy to get up to speed on, although I always thought CouchDB's API was very straight-forward, particularly if you're familiar with HTTP already.

If it were something I was seriously thinking about for production, I'm not sure I'd go with MongoDB.

Personally, I think Riak looks like one of the most interesting of the noSQL bunch. You may also want to look at a graph DB such as Neo4j for something different again.

Answer 5

If you don't have a specific problem to solve, I recommend that you pick up a list of a dozen or so "NoSQL" databases and try them all out and form your own opinions about them. The next time you walk into an interview or a project meeting, rather than being able to say, "I played a bit with FooDB at home", having a qualified opinion on the "NoSQL" phenomenon and a broader overview of the matter is probably going to help you more.