> But isn't it recommended to run the server with fsync? If so, you
> disable it on a benchmark then.
Recommended for what? If you have a system that has to be up 24/7, then you
1) Sound, working hardware
3) Failovers / RAID
(you don't get a hardware failure, OR you have implemented 3)),
(you don't get a power cut OR you have implemented 2)),
then there is no danger that I can see in having the OS handle your disk
caching. If you have a hardware caching controler, then the whole fsync
issue is slightly more academic, but disabling fsync will generally always
IMHO, if you don't trust your hardware and arrangements enough to feel
happy with switching off fsync, then you shouldn't be using that setup in a
mission critical application. It's OK for a development system where you
expect the system to be trashed several times a week, but I am not sure
that fsync should be high enough on the list of priorities in a mission
crytical system to be worth worrying about too much.
Feel free to disagree...
pgsql-general by date
|Next:||From: Jarmo Paavilainen||Date: 2000-12-29 14:35:08|
|Subject: SV: MySQL and PostgreSQL speed compare|
|Previous:||From: Alfred Perlstein||Date: 2000-12-29 13:51:29|
|Subject: Re: MySQL and PostgreSQL speed compare|