Summary:

Analytics database startup MemSQL has integrated JSON support into its big, fast in-memory SQL database. Bridging both worlds is a compelling idea, although execution isn’t always easy.

MemSQL, the analytic database startup from former Facebook engineers Eric Frenkiel and Nikita Shamgunov, is now a NoSQL startup of sorts. The company has added support for JSON, a structure often generated by web and mobile applications, and which has made technologies such as MongoDB and Couchbase so popular.

MemSQL launched in June 2012 after graduating from the YCombinator program earlier that year, promising real-time SQL without the large-vendor price tag. MemSQL uses a number of techniques to achieve its low latency, including in-memory technology and a process by which SQL code is translated into C++ before it’s processed. MemSQL already boasts a handful of customers that include Morgan Stanley, Zynga and Shutterstock.

The idea of fusing relational data and non-relational data (i.e., SQL and NoSQL) isn’t entirely novel, but doing it in a fast, scalable platform is still a pretty rare and alluring proposition. Many NoSQL databases are capable of handling JSON data, but they’re either not designed for analytic workloads or they’re not particularly well suited to large-scale environments. They almost certainly don’t support structured data and SQL.

Another database startup, Drawn to Scale, had integrated MongoDB support into its Hadoop-based SQL database technology before shutting down earlier this year.

The challenge for MemSQL — and any other database company trying to bridge the SQL-NoSQL gap — will be supporting both data types well enough to keep each as a legitimate selling point. There are lots of open source NoSQL databases that do, or can, handle that aspect very well, and there is a whole slew of so-called NewSQL databases trying to rethink SQL like MemSQL is doing.

Most companies would probably love one database to rule them all, but it’s not unheard or particularly difficult these days to manage and connect multiple systems that are each really good at what they do.

Comments have been disabled for this post