dsamperi@Citicorp.COM (Dominick Samperi) (04/12/91)
My organization is considering the use of twisted pair point-to-point connections as an alternative to thin wire Ethernet. The motivation is to reduce our expenses. The environment is one where there is little tolerance for network failures (a trading floor). I'm familiar with thin wire Ethernet, but know little about twisted pair technology. Could somebody comment on their experiences with twisted pair connections. Are they cheaper than thin wire? More/less reliable? Is there a throughput/bandwidth hit in using twisted pair? How large? Is twisted pair easier/harder to maintain? Thanks for any information on this. -- Dominick Samperi -- Citicorp dsamperi@Citicorp.COM uunet!ccorp!dsamperi
henry@zoo.toronto.edu (Henry Spencer) (04/15/91)
In article <1991Apr12.023620.6227@Citicorp.COM> dsamperi@Citicorp.COM (Dominick Samperi) writes: >My organization is considering the use of twisted pair point-to-point >connections as an alternative to thin wire Ethernet... little >tolerance for network failures (a trading floor)... You might want to go have a look at comp.dcom.lans, where 10BaseT (standard twisted-pair Ethernet) has had considerable discussion of late. This isn't really a Unix or TCP/IP issue. (To sum up the c.d.l discussions excessively tersely... 10BaseT works well. Relative costs are somewhat debatable; there is no huge difference, but thinwire may still be somewhat cheaper. 10BaseT is parsecs ahead on reliability for complex networks with large user communities, because its star topology tends to localize failures to a single machine, whereas thinwire takes down a whole network segment when one ignorant clod unplugs or damages a connection.) >Is there a throughput/bandwidth hit in using twisted pair? ... There are slower twisted-pair technologies in use, but 10BaseT is Ethernet in all respects as far as performance goes. -- And the bean-counter replied, | Henry Spencer @ U of Toronto Zoology "beans are more important". | henry@zoo.toronto.edu utzoo!henry
osh@jhereg.osa.com (John M. O'Shaughnessy) (04/16/91)
Twisted Pair networks are likely to cost more in terms of hardware than thin Ethernet networks because they require a concentrator, and you can't daisy chain stations. Twisted Pair 10Base T may save you money in installation costs if the wiring in your facility is up to spec, and if you have enough spare pairs to use for Ethernet (2-pairs). We have found twisted pair networks to be much more reliable due to the terminal-hub nature of the network connection. One station's problems won't affect anyone else. There are also a number of management tools available frm the concentrator manufacturers that help you to manage the network, keeping track of usage, etc. In an area not staffed with technical people, users appreciate the simple telephone cord-like connection as opposed to a cable TV like coaxial connection. I still prefer thin Ethernet for lab areas, or other areas with many machines in close proximity, but for the office environment, we almost always choose twisted pair 10BaseT Ethernet. John -- John M. O'Shaughnessy osh@osa.com Open Systems Architects, Inc. Minneapolis, MN
doug@psy.uwa.oz.au (Doug Robb) (04/17/91)
dsamperi@Citicorp.COM (Dominick Samperi) writes: >My organization is considering the use of twisted pair point-to-point >connections as an alternative to thin wire Ethernet. The motivation is >to reduce our expenses. The environment is one where there is little >tolerance for network failures (a trading floor). I'm familiar with >thin wire Ethernet, but know little about twisted pair technology. >Could somebody comment on their experiences with twisted pair >connections. Are they cheaper than thin wire? More/less reliable? >Is there a throughput/bandwidth hit in using twisted pair? How >large? Is twisted pair easier/harder to maintain? I have recently been to a few seminars where the suggestion has been made that the way to go is to blanket wire with unshielded twisted pair and have a concentrator at some point. With the appropriate cards in the concentrator you can run ethernet, apple talk, token ring etc without having to re-wire the building. Since the seminar's were put on by companies in the market for the above hardware then naturally I'm a bit suspicous as to whether this is any cheaper than using thin wire ethernet in combination with twisted pair as I do here. For a start the concentrator and cards are big dollars.... On the other hand running twisted pair back to a 10baseT hub may be quite cost effective if you are not sure about location of (or want the flexability of moving) serial printers/faxes/terminal etc around the building. The rational would be that you have to run the twisted pair anyway for the above so why not use it for your ethernet devices as well. In the Psychology dep we run twisted pair back to terminal servers, have thick and thin wire coax connecting out mini's, sparcstations and pc's. I think this is the cheapest way to go? Any comments? You could consider running thin wire ethernet and twisted pair? Since thin wire is about $A1.50 a metre you can run one or two segments (up to 180m) on each floor and dont bother with the T connectors etc until you need them. Then for example you want to connect a sparc station you open up the cable tray, fit a faceplate and connector and simply plug in. Since you can hang 29 devices off each segment it seems to me that this is easier than having 29 SEPARATE runs of wire going back to the 10baseT hub rather than the one in the case of the thin wire. Also the mac length of the 10baseT run is 100m , 180 for thin wire and about 500m i think for thick wire. Just to give you an example. I heard of a firm that got 3 floors of a new building in Perth wired up recently. Two thin wire segments on each floor and thick wire segment between floors. No connectors terminators, delni, desta etc because as yet the don't have a computer system. The cost for this $A7,000. They also had 8 pair, PDS (unshielded twisted pair) to 100 outlets with rj45 connectors back to a distrib board put in at the same time, cost $A48,000. To actually get a computer network will need a 10baseT hub etc as you know. Since they don't have a network yet I don't know that the final cost of each scenario would be. doug@psy.uwa.oz.au
andrew@jhereg.osa.com (Andrew C. Esh) (04/18/91)
Just a suggestion, but wouldn't this subject work a little better in comp.dcom.lans? The discussion there is about cable and boxes. My impression of comp.protocols.tcp-ip is that it is above the physical layer of the OSI stack. -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Mpls, MN 55416-1528 Punch down, turn around, do a little crimpin' (612) 525-0000 Punch down, turn around, plug it in and go ...
ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) (04/18/91)
andrew@jhereg.osa.com (Andrew C. Esh) writes: >Just a suggestion, but wouldn't this subject work a little better in >comp.dcom.lans? The discussion there is about cable and boxes. My >impression of comp.protocols.tcp-ip is that it is above the physical layer >of the OSI stack. In fact, a discussion about relative merits of thin Ethernet and 10BaseT has been going on there for a couple of weeks. I'm afraid I caused it by my question 8-() Very true, that doesn't necessarily have much to do with TCP/IP. I've gathered some of the responses - will be glad to send them on request. Please use mail, not a followup to this group. E. -- Eric Behr, Illinois State University, Mathematics Department Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu
GBODSO1@NUSVM.BITNET (HC Eng) (04/20/91)
>Just a suggestion, but wouldn't this subject work a little better in >comp.dcom.lans? The discussion there is about cable and boxes. My >impression of comp.protocols.tcp-ip is that it is above the physical layer >of the OSI stack. Can someone tell me how to subscribe to lists with names separated by periods, like comp.dcom.lans? I am on BITNET only. From the above, am I right to say that the list which I am now reading from, which I subscribed to by sending a message to LISTSERV at UTDALLLAS, and is now distributed by TCP-IP@NIC.DDN.MIL, is also called comp.protocols.tcp-ip somewhere else? HC Eng - Singapore
willis@cs.tamu.edu (Willis Marti) (05/01/91)
Doug Rob (doug@psy.uwa.oz.au) writes:
"Just to give you an example. I heard of a firm that got 3 floors
of a new building in Perth wired up recently. Two thin wire segments
on each floor and thick wire segment between floors. No connectors
terminators, delni, desta etc because as yet the don't have a computer
system. The cost for this $A7,000.
They also had 8 pair, PDS (unshielded twisted pair) to 100 outlets
with rj45 connectors back to a distrib board put in at the
same time, cost $A48,000. To actually get a computer network
will need a 10baseT hub etc as you know.
Since they don't have a network yet I don't know that the final
cost of each scenario would be."
Although one might question the costs, let's also point out that the above
comparison is definitely apples and oranges. Why?
Two thinwire segments = 60 machines, including the required bridge or
(*ugh*) repeaters. You probably meant 8-conductor, or 4 pair, times 100
gives 100 data *and* 100 phone outlets. If you want, you can even make it
200 data outlets.
Cost of making it a network:
Thinwire: (2) bridges, plus 3 sets of terminators. And transceivers.
UTP : Hub
Cost of installing machines 1-59:
Thinwire: Labor and material to cut, reterminate, and place the thinwire
segment, plus network downtime.
UTP: 1/100th hub cost plus plugging a jumper into the wall.
Cost of installing machines 60+:
Thinwire: ???? {new bridge/repeater}, new cable, etc.
UTP: 1/100th hub cost plus plugging a jumper into the wall.
-----------
I believe thinwire can be cost effective in limited settings, but not in a
larger arena (unless you want to saddle your successor with the problems.)
Cheers,
-------------------------------------------------------------------------------
Willis F. Marti Internet: willis@cs.tamu.edu
Director, Computer Services Group, Dept of Computer Science, Texas A&M Univ.
---Not an official document of Texas A&M---