Last month Google released a paper on its high availability datastore Megastore. Megastore “blends the scalability of a NoSQL datastore with the convenience of a traditional RDBMS in a novel way, and provides both strong consistency guarantees and high availability,” the paper says. Megastore is the technology behind Google’s High Replication Datastore, which covered here previously.
It’s a short paper, only 12 pages long. But in case you want something quicker, here are two summaries:
For an even quicker overview, here are three key points:
- It has been widely deployed for internal use at Google for several years.
- It uses the Paxos algorithm to manage replication between data centers.
- Although it offers an eventual consistency mode, it can instead prioritize consistency over performance. It offers three levels of read consistency: current, snapshot and inconsistent.
At the moment, this is only available to the public through App Engine. Perhaps Google will open source it eventually?
