Ex-Facebookers launch MemSQL to make your database fly
With Facebook engineers, it appears the high-performance database apple doesn't far fall from the tree. On Monday, former Facebookers Eric Frenkiel and Nikita Shamgunov (who also spent six years as a senior engineer on Microsoft SQL Server) …
Context & Ripple Effects
MemSQL's debut lands midway through a decade-long argument about databases: the NoSQL backlash of 2009 splintered the market away from relational systems, yet the webscale operators who drove that exodus kept hitting walls — GigaOM reported in 2011 that Facebook was stuck with a MySQL setup its own engineers called a 'fate worse than death'.
The founders embody both sides of that arc: Eric Frenkiel comes out of Facebook's infrastructure grind, while Nikita Shamgunov spent six years as a senior engineer on Microsoft SQL Server. Their bet is that cheap RAM lets a startup offer SQL semantics at webscale speed, and the trade press took notice immediately — DBMS2 ran the story twice on launch day.
First-order effects
- High-throughput web operators weighing MySQL sharding or NoSQL rewrites gain a new option: a from-scratch in-memory relational engine pitched by people who lived the scaling problem at Facebook.
- Shamgunov's SQL Server pedigree puts the startup in direct rhetorical competition with Microsoft and Oracle's installed bases, not just the NoSQL newcomers.
Second-order effects
- Incumbent relational vendors and NoSQL projects alike face pressure to match in-memory performance claims, since MemSQL's pitch attacks both camps' weaknesses at once — SQL compatibility and raw speed.
- Alumni of hyper-scale operators become a recruiting currency in the database market, as buyers read Facebook-scale experience as validation that a young engine can survive production loads.
Third-order effects
- If falling RAM prices keep outrunning disk economics, the SQL-versus-NoSQL debate of 2009 could collapse into a question of storage engine choice, with relational interfaces restored as the default even at web scale.
- A durable pattern forms: infrastructure startups founded by veterans of the biggest web operators, selling back to the industry the internal tools those operators were forced to build.
The trend: Databases are swinging back toward SQL as in-memory hardware makes relational engines viable at the scale that once demanded NoSQL, with webscale-operator alumni leading the charge.