Re: Experimental ARC implementation

Jan Wieck wrote:
> Bruce Momjian wrote:
> 
> > Jan Wieck wrote:
> 
> >> What doing frequent fdatasync/fsync during a constant ongoing checkpoint 
> >> will cause is to significantly lower the physical write storm happening 
> >> at the sync(), which is causing huge problems right now.
> > 
> > I don't see that frankly because sync() is syncing everying on that
> > machine, including other file systems.  Reducing our own load from sync
> > will not help with other applications writing to drives.
> 
> You have 4 kids, Bruce. If you buy only two lollypops, how many of them 
> can share the room unattended?
> 
> What I described is absolutely sufficient for a dedicated DB server. We 
> will be able to coordinate the resources between the various components 
> of PostgreSQL, no doubt. Everyone who has significant performance 
> problems because of I/O saturation, and is still keeping other I/O heavy 
> applications on the same box instead of separating the things, is either 
> not serious or dumb ... or both.

Yes, but my point is two-fold  --- first, someone reported we can do
open, fsync, close reliably, second, I need that for Win32 (no sync),
and third, if we can handle the case for servers with other applications
on the box, why not do that?

--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
359-1001+  If your life is a hard drive,     |  13 Roberts Road +  Christ can be your backup.        |  Newtown Square,
Pennsylvania19073