Path: utzoo!attcan!uunet!cs.utexas.edu!ut-emx!spurgeon
From: spurgeon@ut-emx.UUCP (Charles Spurgeon)
Newsgroups: comp.dcom.lans
Subject: Re: How long can you run ethernet?
Message-ID: <31639@ut-emx.UUCP>
Date: 14 Jun 90 14:21:02 GMT
References: <90164.140927KDM101@psuvm.psu.edu>
Reply-To: spurgeon@emx.UUCP (Charles Spurgeon)
Organization: UTexas Computation Center, Austin, Texas
Lines: 57

In article <90164.140927KDM101@psuvm.psu.edu> KDM101@psuvm.psu.edu (Kevin Maher) writes:
>We have several labs here on campus that we would like to interconnect.
>One problem is that the combined length of cable for the 3 labs exceeds
>600 feet (which I have heard to be the limit of ethernet when using coax).
>We estimate the length to be just over 700 feet.  What I would like to know
>is if this would cause any problems, or if the 600 foot limit is just a
>recommended suggestion.  If it would be a problem, what can we do to get
>the 3 networks connected without having to purchase expensive repeater
>boxes? (we're working on a very limited budget here)   

10BASE5 thick Ethernet, using Belden 9880 (PVC) or 89880 (FEP) or
equivalent coax, has a maximum segment length of 500 meters or about
1,640 feet.  Up to 100 stations can be attached, spaced 2.5 meters
(8.2 feet) apart.

10BASE2 thin Ethernet, using RG58A/U or RG58C/U cable or Belden 9907
coax, has a maximum segment length of 185 meters or about 600 feet.
Up to 30 stations can be attached, spaced .5 meters (1.7 feet) apart.

To go beyond these lengths requires a repeater or bridge.  So if
you're using thick Ethernet you're ok.  I'll assume you're using thin
Ethernet (from the evidence of the length you cite above) and go on to
mention some problems you could encounter.

Overlong segments can result in excessive round trip times for packet
propagation, which could lead to the failure of the Ethernet collision
detect mechanism.  On a sufficiently busy network, the result can be
so many packet failures that users report a "slow" network.

Overlong segments can also have too high a signal loss, resulting in
"ghost hosts" that appear and disappear on the network.  In other
words, the signal level gets so marginal that, depending on the number
of active hosts, packet traffic, and phase of the moon, some hosts can
be reached some times and not other times.  I've seen this happen on
several overlong 10BASE2 thin Ethernet systems in the past.

Back in the dark ages the original 3COM EtherSeries thin Ethernet
system (the precursor to the IEEE 802.3 10BASE2 thin Ethernet system)
allowed up to 300 meter segments, and when people started using
10BASE2 equipment on those networks the performance could get real
strange.

If you're using thin Ethernet, the least expensive repeater I know of
off hand (and have the pricing info on) is the Cabletron MR2000, list
price $865.00.  It comes with one thin Ethernet port and one thick
Ethernet port.  You can attach the thick Ethernet port to another thin
Ethernet cable with a Cabletron ST500-03 BNC equipped transceiver,
list $275.00.  That means you've used up all the ports to link the
thin Ethernets together, and do not have an attachment to the building
network, if there is one.  

There's also the MR5000, a 4 port thin Ethernet repeater with a thick
Ethernet port for attachment to the backbone network.  This lets you
interconnect up to four thin Ethernet segments.  At a list of $2195,
that works out to $439 per Ethernet attachment.  Educational sites can
often get good discounts on these prices.  Cabletron's phone number is
(603) 332-9400.