* Myron Scott <mscott(at)sacadia(dot)com> [010102 08:47] wrote:
> Alfred Perlstein wrote:
> > It's possible what you're seeing is the entire process
> > wait for a disk IO to complete.
> > I'm wondering, how many lwps does your system use? Are all
> > the threads bound to a single lwp or do you let the threads
> > manager handle this all for you?
> Yeah, I looked at this. I have one thread per process that
> does all flushing of buffer pages at transaction commit.
> The client threads register buffer writes with this thread and
> wait for this thread to complete writes
> at transaction end. Unfortuately, selects also wait which really
> isn't nessessary. I hoped this would speed
> simultaneous connections. I created this as both
> a bound thread with its own lwp and a threads manager
> managed thread. I eventually settled on a threads manager
> managed thread, thinking that I wanted to set the priority
> of this thread low and commit as many transactions as
> possible simulateously. Maybe, I should rethink this.
> As for client threads, that is managed by TAO and I
> haven't really played with that configuration.
I think you're saying you have an lwp in what you used to
be per-process context? Meaning if you have 4 connections,
you should have 4 lwps doing disk IO?
If that's true, most likely your problems may be elsewhere,
but since you seem to be overcaching your writes you may
be pessimizing the writes. I wouldn't try to buffer it all,
instead I would hold off until a write is done that isn't
contiguous to the last write position and then being a flush
-Alfred Perlstein - [bright(at)wintelcom(dot)net|alfred(at)freebsd(dot)org]
"I have the heart of a child; I keep it in a jar on my desk."
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2001-01-02 17:48:56|
|Subject: Re: [HACKERS] Re: Inheritance docs error. |
|Previous:||From: Peter Eisentraut||Date: 2001-01-02 17:14:01|
|Subject: NetBSD libedit (Re: GNU readline and BSD license)|