PLus Net problems... (slightly OT)

[Ian Lowe <ian@xxxxxxx>]

Are not really Plus Net problems after all. I am cracking up on this
one, and any suggestions would be welcome.

Our Home Network:

We have a server (turing) with an ISDN card fitted (a BT Speedway/AVM
Fritz PCI) This card is presented to NT as a network card via an NDIS
driver, rather than as a modem.

This runs WinRoute Pro, a NAT firewall package, and connects to various
ISPs.

The problem:

Bizzare packet delays. If I run Unreal Tourney, I get a ping on around
100ms to a fast server. If >anything< else tries to use the internet
connection, the ping rate rises, not to the normalish 200-250ms that I
would expect, but to a freakishly high 3-4000ms.

This originally appeared when we both tried to play UT online, however,
I can cause the same effect using a command window and a web browser.
in command line, I ping a server with -t option (goes forever) then
open
a web page. ping times rise to 1900+, and packets time out, whenever
web page completes, pings return to normal.

This is a recent problem. We have always had some interference when
myself and Jenni play online, which I had put down to winroute, however
it has become ridiculous in recent times. (since early december)

Any single PC however is easily capable of downloading at 7K+ per
second, and Bandwidth testing via CNET consistently gives 60+kbps

History of Troubleshooting Steps:

1) Determine whether problem is related to Plus NET using Surftime.
Plus NET provided a direct dial, non-surftime, number with the same
result therefore BT's "surfport" network is not to blame.

2) Determine whether Plus Net's core network is the cause. Switch to
Freeserve, who use Energis as their carrier. Same problem exists,
suggesting that Plus NET are blameless in this problem.

3) Rule out Winroute Pro. I switched the ISDN Card across into
Feynman(a Win2K Server). Using Windows 2000 RRAS with NAT enabled, I
set up an interface to plus net. The same problem occurs, suggesting
that Winroute Pro (or for that matter, the NT4 IP stack) is not a
factor.

4) Determine whether cable length/physical locality is an issue. This
has been worse since we moved offices (we used to use our small
"study"
as an office, we now use bedroom 2), so cabling could be a factor. I
moved Turing to right beside the Home Highway box, and cabled it using
the manufacturer supplied cable. The problem did not diminish at all.

5) Determine whether any NAT solution will cause this problem. I
disabled Winroute, and performed the same ping/web test from Turing
itself. The problem happens in this test also, which I >think< clears
my entire home network of blame.

6) BT have diagnosed my Home Highway installation, and it performs
correctly. I cannot say for sure, but it seems okay.

7) My Speedway ISDN Card may have been fragged in the same power spike
that killed two of the other machines: Tesla(The TV/DVD/MP3 Station)
and Babbage(My wife's PC).

I am unsure how to proceed from here. As I said earlier, high traffic
is fine. It seems that small ICMP packets like pings don't affect the
throughput (for instance, starting four or five ping windows does not
affect the response times at all) similarly big "chunky"
downloads from
a single site don't seem to have an impact. I could play UT quite
happily whilst D/L a new map pack, without problems (perhaps 220 ping
rather than 130).

It seems, on the face of it to be a combination of a large number of
packets with a reasonable data load which causes the problem. Or it
could be UDP. Do web servers return pages by UDP or TCP?

Help please, before I take to the kids next door with a bit of 2x2 and
a nail.

:)

Ian.


=====
>+++
Ian R Lowe. Director, Wintermute Consultancy Ltd.
e-mail: ian@xxxxxxx
Onward and Upward! Towards Ascension!

__________________________________________________
Do You Yahoo!?
Get email at your own domain with Yahoo! Mail.
http://personal.mail.yahoo.com/