On Tue, Oct 28, 2003 at 01:03:58AM -0800, Mike Stump wrote:
> > > Connection attempts  1193
> > > Connection successes 1047
> > > Consecutive failed connections 0
> > > Probability of connection failure 0.017196883744822623
> > 
> > Umm, why is this bad?
> 
> On my node, connectingTime is in the 10s range.  The above was doing 1
> a second.  If connection bring ups are expensive (I thought they
> were), then firing up 1 a second to a node we were just connected to 1
> second ago, seems, well, odd to me.  If we compute the bandwidth used,
> surely the other node would be less loaded (in terms of CPU used and
> upstream used) leaving the connection up.

Okay so your uptime was about 20 mins?
> 
> The # of open connection is about as free as things get in freenet.
> The don't kill os with context switches, they don't take lots of
> memory, unlike threads.
> 
> No connection should be reopened within 5x the connecting time, or put
> another way, leave connections open if they haven't been idle for 5x
> the connecting time.  If not 5x, how about 1x?

Multiplexing should deal with most of these issues...
> 
> > How do you know they are in your store?
> 
> I was mistaken, it was a new edition.  I was surprised as it hadn't
> been updated in a while and then it was.
> 
> > > messageSendTime is curious, 3 orders of magnitude improvement in the
> > > maximum.  Stunning.  Now, it remains to be see if that is due to
> > > improvements or lack of load on my node.  [shakes head] 3 orders of
> > > magnitude.
> 
> Nope, didn't hold.  I've been hammer back into the ground, slightly
> surpassing the old numbers. 1.4E7 max for messageSendTime.  :-( Could
> we limit inbound requests once this rises above 60s and upstream > 75%.
> 1.4E7 just seems pointless.

We do. After 1000ms.
> 
> > > Right now, these are using up all the job slots and stalling out my
> > > outbound connection even though I have 9 transmitting.
> > 
> > Eh? They will be using up threads, you mean it's queryrejecting because
> > of threads?
> 
> Yes.

-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to