[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Weird slow network problem...



On 24 Aug 1999, Jens Ritter wrote:
> Dan Everton <z1159684@cit.gu.edu.au> writes:
> 
> > The weirdest part about this is that he has identical network card
> > (RTL-8029, using the ne2k-pci driver), identical kernel (2.2.11), identical
> > Debian install (potato as of yesterday) and identical config (except for the
> > IP address of course :). I have no problems transferring files at full
> > speed both locally and externally, he does.
> 
> Interrupts and other cards are the same, too?
> 
> What do you mean by locally and externally? Two cards in different
> directions?
> 
> Jens

No interrupts and other cards are not the same. I have an ISA VIBRA16 and
he has an ISA ESS688, I have an PCI S3 ViRGE, he has a PCI Tseng ET6000, I
have a Voodoo2, he doesn't. It's not a hardware problem as far as I can tell
because it only affects external links not local ones.

As for what I mean by locally and externally... we both live on campus at
university and are connected to the residential network. Traffic within
this network is fine and goes at full speed for my friend. The residential
network is connected (through a firewall) to the university network.
Traffic to this network is extremely slow for my friend but normal
ethernet speeds for me. I hope that clears it up.

The other weird aspect of this problem is that transfers made by a machine
outside the residential network to his machine go at full ethernet speed
as well. For example, using ftp  on an machine outside the
residential network to copy a file to his machine goes at about 200kB/s.
Using ftp to copy a file from his machine to a machine outside the
residential network goes in ~30kB/s bursts with a few seconds between
bursts. I get roughly same speed (200kB/s) in both directions. If it helps
the external machine we're connecting to is a Sun box running SunOS 5.6.

Oh, and before you think it's an ftp problem, this affects all other
protocols (including ICMP if the bing results are any indication). And
there's no measurable packet loss.

This one's been puzzling us for a while so I hope someone out there has a
solution.

Dan



Reply to: