Spanner: Google's globally distributed database

James C. Corbett, Jeffrey Dean, Michael Epstein, Andrew Fikes, Christopher Frost, J. J. Furman, Sanjay Ghemawat, Andrey Gubarev, Christopher Heiser, Peter Hochschild, Wilson C. Hsieh, Sebastian Kanthak, Eugene Kogan, Hongyi Li, Alexander Lloyd, Sergey Melnik, David Mwaura, David Nagle, Sean Quinlan, Rajesh Rao, Lindsay Rolig, Yasushi Saito, Michal Szymaniak, Christopher Taylor, Ruth Wang, Dale Woodford
[doi] [Google Scholar] [DBLP] [Citeseer] [url]
Read: 21 September 2020

ACM Trans. Comput. Syst. 31(3)
Association for Computing Machinery
New York, NY, USA
August 2013
Note(s): google, Paxos
Papers: chang:tocs:2012

Spanner is a scalable, globally-distributed, temporal database that shards data across many Paxos instances. It features automatic failover, resharding and migration. Although not explicitly stated, it seems to be a replacement for Bigtable (chang:tocs:2012) and Megastore. One of the early clients was to support F1: an ads database.

Spanner consists of something close to a temporal relational database, layered on top of a temporal key-value store, made robust against failure using Paxos and relying on “TrueTime” to keep global datacenters synchronized and allow global serialization of all transactions.

“TrueTime” is a time service that provides time intervals that are guaranteed to be globally valid. This is implemented using a combination of GPS and atomic clocks. The atomic clocks are referred to as “Armageddon masters” and serve mostly as a backup in case GPS fails.

The paper puts a lot of emphasis on the importance of TrueTime

As a community, we should no longer depend on loosely synchronized clocks and weak time APIs in designing distributed algorithms.

  • Why Google stores billions of lines of code in a single repository [potvin:cacm:2016]