Yesterday, Google Research pulled the shroud off Spanner, Google’s “scalable, multiversion, globally distributed, and synchronously replicated database,” claiming to have created “the first system to distribute data at global scale and support externally consistent distributed transactions.”
The only problem?
TransLattice has been deploying a globally available, geographically distributed multinode database for almost two years.
I wrote about TransLattice almost two months ago when the company finally went public. The company’s core product is “TED,” TransLattice Elastic Database, and it enables a live, functional, operational database with multiple nodes all over the world, without simply syncing, replicating, or cloning.
Spanner, which appears to be in use at Google, maintains global consistency with the assistance of atomic clocks and GPS. The database automatically places data closest to where it’s frequently needed, replicates elsewhere for redundancy and backup, and can scale impressively: up to “millions of machines across hundreds of datacenters and trillions of database rows.”
That sounds like greater scalability than TransLattice, which two months ago was running on up to 12 nodes and was working on capability for 20 or more nodes.
But TransLattice believes it was first, saying that Google’s system, while interesting, is much more a key-value store than a true relational database, which while enabling huge scale, would not work for enterprise applications.
In other words: It’s built for exactly what Google does every day.
In addition, however, TransLattice co-founder and chief technical officer Mike Lyle called Google’s claim to be first into question.
“Most dubious is Spanner’s claim to be the first system with these properties. TransLattice has been delivering to customers true SQL conformance and geographically distributed relational databases with all of the ACID guarantees for nearly two years.”
I talked to a Google representative about Spanner and TransLattice, and while she did not have a specific quote, she said that Spanner is a Google Research project, not a specific product.
Any specific claims in the paper are not claims Google per se is making, and while Spanner is deployed right now, she could not comment on how long it has been operational.
Reading through the Google Research paper (OK, skimming it) it appears that Spanner has been built for a different use-case than TED. And both, to be completely frank, are incredible achievements.
No matter which one happened to be first.
photo credit: IronRodArt – Royce Bair (NightScapes on Thursdays) via photo pin cc