Spanner (database)

from Wikipedia, the free encyclopedia
Stretcher
Basic data

developer Google Inc.
category Database
cloud.google.com/spanner/

Spanner is a globally-distributed NewSQL - database of Google Inc. release is the successor to the Google databases Bigtable and Megastore .

Spanner according to the classification by the CAP Theorem consistent ( consistent , C ) and partition Tolerant ( partition tolerant , P ), but also has a very high uptime of about 99.99958% and is therefore de facto also available ( available , A ).

Spanner is publicly available as a paid cloud application. The database management system (DBMS) Google F1 , which was previously based on Google's own MySQL variant, is also based on Spanner.

functionality

Spanner uses a relational data model . However, every table must have an index column. Physically, the data is stored append-only as an " interleaved table " (analogously: "nested table"). Competing access to data records is implemented using Multiversion Concurrency Control (MVCC). Records from different tables are alternately stored in a single list and distributed over several shards .

Time synchronization takes place using Google's TrueTime infrastructure, which uses atomic clocks and high-precision GPS receivers in every data center for time synchronization and thus guarantees globally valid time stamps. The time stamps consist of two times, one guaranteed to be in the past and another guaranteed to be in the future.

In the case of a write access, the record to be written is provided with a TrueTime time stamp as part of a transaction in order to be able to guarantee a uniform sequence of write accesses across all shards. The write accesses are synchronized over several shards using Google's Paxos algorithm and then waited for a specified time at which the write access should take place before the record persists synchronously and the transaction is completed.

A time stamp is also given for queries so that the results of distributed MapReduce queries are consistent. This also enables historical views of the database.

Web links

  • Cloud spanner. No-Compromise Relational Database Service. In: Google Cloud Platform. Google, accessed March 21, 2017 .

credentials

  1. ↑ a b c tensioner. Google's Globally-Distributed Database. Google Research, September 2012, accessed on March 21, 2017 (2012 white paper).
  2. ↑ Eric Larson: Google's Spanner: Database Tech That Can Scan the Planet. In: Mashable. October 21, 2012, accessed March 21, 2017 .
  3. ↑ a b Spanner, TrueTime and the CAP Theorem. Google Research, February 14, 2017, accessed March 21, 2017 .
  4. ^ A b Cade Metz: Spanner, the Google Database That Mastered Time, Is Now Open to Everyone. In: Wired. February 14, 2017, accessed March 21, 2017 .
  5. ↑ a b c d Sebastian Kanthak: Spanner. Google's Distributed Database. InfoQ, November 28, 2013, accessed on March 22, 2017 .