Linear elastic caching reduced Spanner's memory use by 15.5%(research.google)

https://research.google/blog/optimizing-cloud-economics-with-linear-elastic-caching/