Problems with network performance.

Mike Brodbelt m.brodbelt@acu.ac.uk
Tue Apr 27 09:16:16 1999


Donald Becker wrote:
> 
> On Mon, 26 Apr 1999, Mike Brodbelt wrote:
> 

> > eth0      Link encap:Ethernet  HWaddr 00:00:C0:0C:75:E8
> >           RX packets:18457653 errors:1717 dropped:0 overruns:0
> >           TX packets:15145268 errors:79182 dropped:0 overruns:0
> 
> You should be reporting the statistics from /proc/net/dev, rather than
> summarized numbers from 'ifconfig'.  But I can tell immediately that
> something is broken, almost certainly your 10base2 cabling.
> 
> >           RX packets:18831732 errors:1777 dropped:0 overruns:0
> >           TX packets:15463638 errors:81207 dropped:0 overruns:0
> > Errors shot up to 81K during mornings testing.
> 
> Ouch!  Try repeating the test using a dedicated 10base2 network, with no
> breaks in the cable.

Ok, statistics on the same machine from /proc/net/dev are:-

Inter-|   Receive                  |  Transmit
 face |packets errs drop fifo frame|packets errs drop fifo colls
carrier    lo:1788411    0    0    0    0  1788411    0    0    0    
0    0
 dummy: No statistics available.
  eth0:20276190 1928    0    0 3547 16567418 84469    0    0 36591 95386

What does the driver consider a TX error, and how does it differentiate
between an error and a collision?
 
Machines A and B are on the same (short) thinnet segment, and testing
after diconnecting the rest of the network (leaving a test net about 20'
long with around 5 machines attached) showed some interesting results.

FTP transfer of 7093877 byte file:-

Under test net conditions -  6.25 secs, = 1108 kb/s
With whole 10Base2 network - 30.6 secs, =  226 kb/s

This seems to suggest that the problems are pretty much all cabling
related, which at least makes me happy that my linux boxes are behaving
properly, even though I'm less than ecstatic about the state of the
wiring.

The network here consists of about 12 segments of 10Base2 linked with
repeaters. It's pretty much a textbook example of how not to cable a
network, and it's falling apart at the seams - it's in desperate need of
a complete rewiring job.

Anyway, many thanks for the help, I think I now know what the trouble
is. I'd still ilke to know how the Netware server manages 600 kb/s -
does it just break specs and aggressively retransmit, or is this a
difference between IPX and IP?

Mike.