[pptp-server] Maximum number of simultaneous connections
Neale Banks
neale at lowendale.com.au
Sun Jun 17 19:38:45 CDT 2001
On Sun, 17 Jun 2001, Torge Szczepanek wrote:
> Neale Banks wrote:
>
> > That'll most likely be memory limitations.
> >
> > However, there are other practical limitations, as exemplified in this
> > snippet from linux-2.2.19-orig/net/core/dev.c:
> > * If you need over 100 please also fix the algorithm...
>
>
> > FWIW, it has the same <=100 limitation in linux-2.4.5.
>
> Okay. But how to break this limitation of 100 ppp devices? Simply
> increasing the number in the for loop may not be enough, as it is
> written in the kernel. Any suggenstions?
I didn't read anything sinister into it, just that the existing strategy
of a serial serach for the first available free dev# works ok for a small
number[1] of devices but doesn't scale for larger numbers devices (i.e.
increasing the (somewhat arbitrary) limit of 100 would probably "work" but
not perform acceptably).
Think of a file system in need of a block for a file - if each time we
needed a block it was to sequentially search all the blocks on disk (or
even just an index thereof) performance would be most likely *horrible*.
File systems don't do it this way - they maintain a "free list". I'd
guess that for >>100 devices a free list of some kind could be a Good
Thing (the nature of such free list being left as an "exercise for the
reader" ;-).
HTH,
Neale.
[1] the existing kernel code obviously implies that suitable values of
small are < 100.
More information about the pptp-server
mailing list