NoSQL is undergoing a rapid evolution, according to the co-founders of MemSQL, a startup whose eponymous database reduces the time needed to crunch enormous amounts of data by leveraging in-memory technologies.
That proliferation of NoSQL databases, which take a non-relational approach to data (as opposed to SQL, which is meant for managing data in relational database management systems), has led to a fair amount of debate among IT pros.
“It’s easy to get into a NoSQL versus SQL joust,” Eric Frenkiel, a MemSQL co-founder, said during a Skype conversation. “We see ourselves as the alternative to NoSQL, where you get all the speed and all the SQL, and you don’t have to give up the query-ability.”
By placing a relational interface within an in-memory data tier, MemSQL can push queries much faster than MySQL (a video produced by the company shows MemSQL pushing 80,000 queries per second, versus MySQL’s 3,000 queries per second, on the same machine) while maintaining a level of performance. For each unique type of SQL query, MemSQL generates and compiles C++ code, which in turn runs against memory-optimized and lock-free data structures.
And whatever database they choose, organizations need speed. Even relatively small companies wrestle daily with a flood of data from customers, social media, supply chains, and other sources. If those firms have embraced NoSQL as a way to manage and analyze that data, it’s because the format offers several benefits, including the ability to store complex structures within a single database record.
However, NoSQL can also lead to problems. Whereas a traditional SQL database retrains database programmers so they don’t accidentally put unmatched data into a table, NoSQL systems such as MongoDB allow the dropping of any type of data into any collection. Accidents can happen as a result.
“There is a use case for NoSQL,” Frenkiel said. “If you’re doing anything with processing, e-commerce, geospatial coordinates” then SQL databases become necessary for their structuring abilities.
Along with Nikita Shamgunov, Frenkiel founded MemSQL in 2011. Both had worked on Facebook before deciding to embark on their own project. While the world’s largest social network offered a world-class challenge in terms of maintaining a database, the two wanted to focus on building their own.
“I really enjoy building systems,” Shamgunov said. “Big memory is a hot category right now, and myself and my friends are experts in OLTP [online transaction processing]. Then we got Y Combinator [which provides startup funding], and those things added up to point us in this direction.”
One thing the pair pulled from Facebook: the idea of offering a free product, then monetizing separate layers atop it. “We want more people to deploy MemSQL,” Frenkiel said. “”We’re going to be monetizing in the cloud. I’d expect to see some partnerships around some cloud offerings; beyond that we do have an enterprise edition.” The enterprise, he added, is in dire need of databases that can crunch information quickly: “They need speed microsecond by microsecond.”