Greetings Sun Managers,
My original message asked for experiences with existing twisted-pair
(TP) wiring of unknown quality in Ethernet networks.
Our wiring has no markings on the insulation to give a clue as to its
type. It looked as if it was untwisted. One response suggested I
check again because wire with few twists per foot may look untwisted
unless you look at a sizeable run of it. After pulling out several feet
of our wire for closer inspection we found that it is indeed twisted.
(An aside: While searching for sources of networking equipment and
services I came across an advertisment that had a picture of a guy and
a girl, both punk/biker types with spiked hair, covered in tattoos and
holding a snake. The caption on the ad read: "If this is your idea of a
twisted pair, call us before installing your network").
Most responses said that using flat ribbon wire (like consumer phone
cables) was unworkable or only usable over very short distances without
massive retries due to collisions.
A few had good luck with this type of cable.
Paulo Licio de Geus (paulo@dcc.unicamp.br) wrote:
"For short distances, you can use plain telephone cables, like the ones
that come with consumer telephones (not twisted, not shielded). For
anything above 20m I would suggest shielded and twisted cables, just
to be on the safe side, and unless further testing is done. We have
tested the consumer phone cables with distances of 15-20 m,
starting/stopping hand drills and vacuum cleaners with several turns
of cable around them. netstat wouldn't record any errors (input and
output)."
Others didn't have much luck, particularly if the existing wire went
through junction boxes, punch down blocks, or other devices instead
of being in one unbroken run to the hub.
sjh@math.purdue.edu (Steve Holmes) wrote:
"Our experience is mixed. It depended, for us, on the number of
connections the circuit goes through from one end to the other. In our
building the phone wiring goes from the office to a group box to the
floor wiring closet. Some of the circuits had more connections than
others, and there was no consistency in the integrity of the lines. Our
decisionn in the end was that we could not trust the house wiring to
work for TPE. So we pull new real (now level 5) 4-pair cable to each
new worstation site."
On the question of using existing twisted pair wiring of unknown grade
the responses were about 50-50.
Half said that it was OK if you used line testers to insure that you
had good wires. The quality of the hub and transceiver (for systems
that don't have them built in) are also factors here.
The other half said don't take chances- pull new wire. Many cited nagging
problems with existing wire that made it worth the cost of pulling new wire.
Another advantage of pulling new wire is that you can use wire that will
work with faster networks in the future.
The current standards were formulated for level 3 wire- a common type
of unshielded twisted pair (UTP) phone wire. Level 5 wire is not that
much more expensive and will support FDDI/CDDI speeds when the hardware
becomes available for twisted pair.
dbrillha@dave.mis.semi.harris.com (Dave Brillhart) wrote:
"Well, I just attended a meeting where we discussed CDDI. It is a
100Mbps protocol over UTP. It requires Catagory-5 TP, which, amoung
other things, is 1 twist per 1/2 inch. They said they guarentee 100'
with CAT-5 TP, and only 40' with CAT-4 and no guarentee with CAT-3.
That all leads me to assume that the twists are important. I imagine
it might work @ 10Mbps for short lengths untwisted, but don't count on it."
Many thanks to the following:
paulo@dcc.unicamp.br (Paulo Licio de Geus)
jrose@mtgy.gtegsc.com (Joe Rose (4710))
john@oncology.uthscsa.edu (John Justin Hough)
rolandp@sv.gpsemi.com (Paul Roland)
Johnson Lew <c3314jcl@mercury.nwac.sea06.navy.mil>
helios!electron.var.com!scb@uunet.UU.NET (Steve Brown)
ebumfr@anah.ericsson.com (Mike Rembis 66520)
etnsed!xhaque@uunet.UU.NET (Amanul Haque)
grs@claircom.com (Gregg Siegfried)
strombrg@hydra.acs.uci.edu (Dan Stromberg)
dbrillha@dave.mis.semi.harris.com (Dave Brillhart)
sjh@math.purdue.edu (Steve Holmes)
amick@erim.org (Richard Amick)
Christian Lawrence <cal@soac.bellcore.com
bert@penril.com (Bert Robbins)
gdmr@dcs.ed.ac.uk (George D M Ross)
szh@zcon.com (Syed Zaeem Hosain)
Paul Begley <peb@sandoz>
wallen@cogsci.UCSD.EDU (Mark R. Wallen)
rheaton@synaptics.com (Rick Heaton)
Jeff Wolfe <wolfe@essc.psu.edu>
Matt_Mauss_-T@hqdev.3mail.3com.com (Matt Mauss)
steve@seattle.avcom.com (Steve Lee)
Regards,
Rick
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:58 CDT