Thanks to everyone who replied to my posting.
We were suffering from chronic poor response and occassional network
hangs.
The problem seems to be cured (through serendipity rather
than skill and deduction). It looks as though we had some faulty UTP
splitters, which were being used to drive two devices off one line.
Since we removed these devices the errors have disappeared and collision
rates have dropped dramatically.
In fact I think we have several problems - only the most serious of which
has disappeared.
The responses:
. Several people suggested upgrading to 2.3. A lot of problems
like those we were experiencing go away with the upgrade.
>
> . What is a reasonable collision rate?
>
The consensus was to keep it <5% (some said 1% others 10%). Nobody
specified that this means average not peak. One respondent said to
aim for 1% av and 2-3 % peak..
> . What is a reasonable error rate?
>
Nil. Some replies said <5% or <1%, but the consensus seems to be that
any but occassional errors mean you have a problem.
It's better to sort it out before it leads to serious trouble.
> . What tools are available to help isolate these problems?
>
netstat, etherfind and snoop were mentioned.
Some people use dedicated network analyzers.
tcpdump and Network General Sniffer were mentioned as FTPable
packages that would help. I haven't followed these up beyond reading the
man pages.
Bob
----------------------------------------------------------------------------
Thanks to:
aw@de.sel.bk (Armin Weber)
Mike Cross <cross@com.ge.med.engfs>
eric.deschamps@fr.diva (Eric Deschamps)
eckhard@de.dlr.go.ts (Eckhard Rueggeberg)
bobr@COM.SLB.wireline.houston ( Bob Reardon )
jkays@edu.msc
Juergen Rothenanger <unrz57@de.uni-erlangen.rrze.rzsuna>
cygan@com.wpm (Linda Cygan)
Mike Raffety <mike_raffety@com.swissbank.us.il>
cecilp@com.rogers.cantel.vancouver (Cecil Pang)
[my mailer has reversed the addresses.]
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:55 CDT