North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Satellite latency

  • From: Richard A Steenbergen
  • Date: Tue Mar 05 19:48:43 2002

On Tue, Mar 05, 2002 at 03:52:31PM -0800, Clayton Fiske wrote:
> > kernel buffer and will misreport speed). Open a few more connections like
> > that and you've exausted your kernel memory and most likely will have a
> > panic. If you did these settings on a web server, all it would take is a
> > few dialups trying to download a big file before you go boom.
> Since this buffer is determined by the receive window (and thus by the
> receiving system), I don't see why it matters. Most receivers will be
> using OS defaults, and the ones who adjust their receive window size up
> are probably doing so because they have a fast enough link to warrant
> it. I can't see where a dialup user would find any cause to give himself
> a huge receive window.

No, the buffer is determined by the configured socket buffer size, the
rate at which TCP drains the socket buffer on the network is determined in
part by the advertised receive window.

So if you run a web server and try to turn up your socket buffers to some
big number thinking you'll get better performance, and a dialup user
connects and tries to download a 1MB file, the web server will immediately
dump 1MB into the kernel until either the socket buffer or the file runs
out, and then the kernel will spend the 5 minutes transfering it to the
dialup user. Have that happen a few times, and you get an instant mbuf
exaustion (or whatever internal mechanism your OS of choice uses) and
kernel panic...

Richard A Steenbergen <[email protected]>
PGP Key ID: 0x138EA177  (67 29 D7 BC E8 18 3E DA  B2 46 B3 D8 14 36 FE B6)