Google Spanner


If Google research article is correct, Google Spanner database is an absolute killer.

  • SQL Language
  • Clustering
  • Two-phase commit with minimal locking done using GPS clock and Paxos cluster consensus protocol
  • Huge tables
  • Column-based storage
  • And all this with a decent support of SQL language.
  • Yes SQL!
  • You heard it right, it supports SQL!
  • Replication of data
  • Automatic balancing replication to make it geographically closer to consumers and thus faster
  • Globally consistent
  • Currently used all around the world
  • Currently used by Google services that are used maybe half a billion users daily.

In my opinion if Google ever releases this database Oracle can close the shop and return money to shareholders. IBM can stop trying to lipstick on DB2 pig. And Microsoft can sit in the corner and cry knowing what a joke Azure is.

But Google is never going to release it. And if someone from open source world starts a project to replicate the functionality of Google Spanner I would love to join this project.

Page Life Expectancy

In SQL Server Buffer Cache Hit Ratio is always at 100%. I never managed to get in lower than it. Instead the metric to look at is Page Life Expectancy. It tells how long in seconds a page is expected to stay in cache. If it drops below 300 you need to increase memory allocated to Buffer Cache.