PostgreSQL pain points

Posted Mar 27, 2014 16:48 UTC (Thu) by Shewmaker (guest, #1126)
Parent article: PostgreSQL pain points

Regarding double buffering, mapping files from hundreds of processes being slow, and madvise--LLNL has developed a new mmap implementation (loadable as a module) that is optimized for data-intensive applications. They also saw slowdowns for many processes using the standard mmap, but theirs scales up to hundreds of processes. Papers and code are available. This is going into production on their clusters.

See DI-MMAP.