5 ms·
You can actually configure postgres to not do an fsync() after each transaction; instead it gets written to disk, which on Linux means written to the disk cache
by patrickgzill 16y ago
You can actually configure postgres to not do an fsync() after each transaction; instead it gets written to disk, which on Linux means written to the disk caches in RAM, then Linux flushes to disk when it does the rest of the disk writes (anywhere from 5 to 30 seconds are set as default on different versions of Linux).
- tedjdziuba 16y agoWhy would you want to do that? This isn't MongoDB, the real world cares about durability.
- patrickgzill 16y agoI am not sure why I was downvoted - I wasn't suggesting as a course of action... it gives a large speed increase and I was suggesting that possibly this was done as a performance measure.