Someone please tell me we’ve gotten past the either-or debate over NoSQL and relational databases.
While NoSQL databases are foundational technologies for web startups — with most of these young companies opting for MongoDB, Cassandra, CouchDB or something else to fulfill their database needs — they might be better served going a hybrid route instead. There’s always room for a good, old-fashioned relational database — especially if they want to conduct and store financial transactions.
Just ask Mark O’Neill, CTO of Thrillist, a New York City-based media company that fields e-commerce and consumer recommendation services. Thrillist uses the NoSQL MongoDB to track and store tons of data about user interactions, but it’s MySQL all the way when it comes to bread-and-butter transactions and financial data that runs the company.
There’s a reason for that, O’Neill said. As great as MongoDB (or Cassandra or CouchDB or insert your favorite NoSQL entry here) may be, they’re still relatively immature compared to their SQL forebears. The ancillary tools aren’t as robust and it’s hard to find NoSQL talent.
“Skillsets around NoSQL are lacking and SQL [as a language] is relatively simple to learn — writing queries in SQL is not so bad. With NoSQL, the tools are less robust and the barrier to entry is much higher,” O’Neill told me in an interview.
Thrillist, founded in 2005, looked at several NoSQL options but went with MongoDB over the NoSQL alternatives because at the time it was more stable, had a larger community around it and better tools than the others, O’Neill said.
Make no mistake: MongoDB is great for handling all the critical social interactions that take place. “For each action taken by a user, you want to know what the user’s friends were doing and you want to pull all that data out from a single location. Say you take an action on Meet Up, it will update your user references and update all your friends. Non-relational stores are really good at that and you can afford to keep that data in multiple places — we use Mongo for that,” O’Neill said.
But, for transactions? Well, “Mongo doesn’t really have transactions. If I write [data] in multiple places and want to check all that in at one time, Mongo can’t do that,” O’Neill said. When someone buys something at Thrillist’s Jackthreads site, the system must record their order and all the items associated with that order, or nothing works. “It all gets written or none of it does. Mongo is not good at that,” said O’Neill.
So a word to the wise web startup: NoSQL — in this case MongoDB — is great for what it does, but for your financial transactions stick with SQL.
For more discussion about the technologies — NoSQL or not — powering the web sites and mobile applications we all use, come check out GigaOM’s Structure conference next month.
Feature photo courtesy of Flickr user takomabibelot
Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.
- A near-term outlook for big data
- Infrastructure Q1: Cloud and big data woo enterprises
- Putting Big Data to Work: Opportunities for Enterprises