Google Spanner and Akiban: Opposite Ends of the Same Spectrum
Google Spanner was mainly built because the most important system at Google, Google AdWords, had accumulated Terabytes of data on MySQL. It is a well known fact that Google has always trusted and relied on a relational database to power the system that generated more than 90% of Google’s revenue. When the amount of data grew, Google developed something called “schematized, semi-relational tables” - or something Akiban calls Table Groups.
This is a big deal: Google and Akiban are using the same approach to storage and we have a similar data model. The difference? We're attacking the same problem from opposite ends of the spectrum. Akiban is currently focused on solid single and multi-node performance while Google is focused on global-scale, Google-sized transactions.
In this presentation, Ori Herrnstadt will present on why a forty-year old technology, Codd’s relational database, is here to stay. He will talk about how the relational database has been evolving and growing with the demands of Big Data. We will focus on Table Groups: why and how they make a difference in the evolution of the relational database. A tool so powerful that Google has developed a storage system exactly like Akiban’s. You will learn how Akiban, in addition to achieving efficiency by co-locating data together, revamped the way queries are executed to take advantage of the entire “object graph” available to us without having to perform any joins. While Google is focused on creating systems that can survive the unexpected destruction of multiple data centers, we’re focused on reapplying some of Codd’s wisdom to what we think has been established as the future direction of relational database storage.