There is similar problem in Postgres as well. Not nearly as bad in magnitude, and with some shared responsibility with Linux and file system code (since PG rel

There is similar problem in Postgres as well. Not nearly as bad in magnitude, and with some shared responsibility with Linux and file system code (since PG relies of filesystem's caching). Greg Smith talked about it in some detail during PGWest 2010. One of his customers had long stalls in production and Greg worked out how to reproduce and explain them. He had some patches he was planning for Postgres 9.1, not sure if they went in.