If you think about it, it’s a bit tragic that the APIs operating systems offer for memory mapped files are not enough for databases. The operating system already has all the caching logic in place and can do so more efficiently with hardware support.<p>If you have data that rarely changes you can actually create an ad-hoc database system based on that and have performance like an in memory database without having to worry much about the memory of the system.
Sounds interesting, I miss a benchmark to go with the article in my opinion.<p>"pg_prewarm makes most sense if your database and your RAM are really really large (XXX GB or more)"<p>A benchmark preferably with databases of sizes that make sense.