[Gllug] Networks

[Jim Bailey]

On Thu, Sep 12, 2002 at 09:59:32AM +0100, Peter Childs wrote:
> 	We are trying to replace an old Bull DPS6 (15 years is old in computer terms 
> is it not) With a linux based system using C++, Qt and postgres. The system 
> controls hire cars travelling round London. If you want more details I will 
> think about giving some. 
> 	We are planning to connect our computers together using a 100Mbits Lan and 
> switches around our offices in SE1. Maximum cable run is about 30 meters 
> which is well below the 100meter limit. However when we try to run data over 
> the links over 8 meters long we get no  data, we get a link (or at least the 
> switch says there is one but no data travels) All the cable works fine at 
> 10Mbits and the configuration seams to work on cables shorter than about 8 
> metres.
Hi,

I am not a networking expert but I have played with networks a bit so
here is my 2cents worth, covered under the GPL. ;)

Have you got packet sniffing software on either end of the run, ethereal
or tcpdump, or even better if the switch is capable on the the switch
itself.  This will help you to diagnose the problem, I hope other people
will have better suggestions.  There is also iftop which shows transfers
across an interface in a format similar to top.

You might also want to check the NICs are running at 100Mb rather than
10Mb. The same could be said for the switch as sometimes
auto-negotiation on some switch/NIC combinations can be problematic.

> 	The main problem we can see is that there might be problems with interference 
> on our runs from UPS, Power (Non UPS), CCTV, Serial Cables (for the bull), 
> phones (4 different types), etc. All the cable is in the same run under the 
> floor.
> 	If any one has any ideas I would love to hear them.
>
I have run dozens of Cat5 cables in runs next to power cables and also
Market Eye live feeds using coax without any noticable effects.  These
ran into fast hubs which were plugged into a layer of switches. some of
the runs were 70 meters plus again without any problems.

If you haven't done it already do a long test run over desks etc away
from power sources to check the interference theory, that should answer
that one.

> Thanks in advance.
> 
> Peter Childs
> PS. I have tried to send this message (or one like it a number of times but it 
> never seams to turn up on gllug!) 
>
It could be symptoms of a wider problem.

Peace Jim

-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug