httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chuck Murcko <ch...@n2k.com>
Subject Re: Patch to slow down children's deaths
Date Wed, 08 Jan 1997 00:53:04 GMT
Ed Korthof liltingly intones:
> 
> 
> On Tue, 7 Jan 1997, Chuck Murcko wrote:
> > Mmmmm. I'm not convinced there is a steady state, since the actual connects
> > are stochastic. You can average, of course, and the average tends to rise
> > & fall rather smoothly over a day, but I see huge & random peaks as well.
> 
> This is exactly why I'd like to decrease the rate of childrens' deaths.
> It seems silly that large numbers will die essentially immediately if
> there's a brief lull; after they're dead, it takes a while to fork them
> again.  High MaxSpareServers and low MinSpareServers will help, of course,
> but the limitations on those are exactly why we have the 1 fork/second
> restriction, no?
> 
Well, the peak handling is also helped by raising MinSpareServers, since
you start 1/sec forking earlier into a peak. That was one of the points I
was trying to make.

Slowing down child deaths is a way to have this behavior in the short term
(whatever the slowdown time is), and sounds desirable for some sites.
Others, though, will need to play the MinSpareServers and MaxSpareServers
game because of the loads they run. I'd think it would be nice to be able
to turn off the slow death if desired.

chuck
Chuck Murcko	N2K Inc.	Wayne PA	chuck@telebase.com
And now, on a lighter note:
When in doubt, do what the President does -- guess.

Mime
View raw message