mep@AQUA.WHOI.EDU (Michael E. Pare) (10/03/89)
You've seen pros and cons of both. The biggest problem with both is in trying to run wiring from the host to the 'backbone' whatever it may be. Thick is a pain if you have several machines in the same area, each requires a drop using bulky transceiver cable, and you can quickly run out of space on the cable (2.5m between transceivers) unless you use a multiport transceiver, but you are still stuck with bulky transceiver cables. An entire backbone of Thinnet is clumsy and can easily lead to wiring faults and hard to trace network problems, and has severe node limitations. One escape may see a thick ethernet backbone with thick-to-thin repeaters used to hook up local groupings of hosts. One repeater can support say 8 areas, with each area supporting several machines (or just one). This provides better fault isolation and enables a large node installation. I would definitely suggest you look into twisted pair format (as someone mentioned). This can be the least costly to install if the twisted pairs (just one using 3COM's system, or two for Synoptics or eventually the 10BASET standard) are already available. You can even install the twisted pairs separately for a lower cost than trying to run a lot of coax. This method provides for the best fault isolation and is the easiest to support if people move around, as well as allowing for a large node installation. I've installed and supported all three and twisted pair wins hands down. By the way, 3COM's twisted pair ethernet is based on thinnet technology while Synoptics is more based on thick.
ian@lassen.wpd.sgi.com (Ian Clements) (10/03/89)
In article <8910022022.AA09420@aqua.whoi.edu>, mep@AQUA.WHOI.EDU (Michael E. Pare) writes: > > ...This can be the least costly to install if the twisted pairs > (just one using 3COM's system, or two for Synoptics or eventually the > 10BASET standard) are already available. I disagree. Remember that you need one transceiver for each workstation regardless of whether or not you use thick, thin or twisted pair (assuming that no workstation already has a thinnet xcvr installed). I'm calling the xcvr costs a wash even though there is a $50 difference between thin and twisted pair (thin being least expensive and twisted pair most expensive). So we can assume that all that is being compared is transmission medium and associated equipment, correct? Following that assumption, the cost to acquire backbone cable for thick or thin costs $1.25/ft and $.29/ft respectivly. Twisted pair cable costs $.05/ft. The twisted pair cable it self is much cheaper however, to make the whole thing work you need this box (commonly refered to as an Active star repeater) that can cost somewhere around 8k for more than 10 connections. Admittedly, once the twisted pair system is up and running, maintenance, additions and other changes are far easier to deal with. One can almost always install more twisted pair cable than either thick or thin backbone and drop cables for the same costs. Here at SGI for example, every office and cube gets a 4 port data block. From that block you can have Ethernet, PhoneNET (AppleTalk) and serial connections. Then of course there is the issue of the 10BASET spec which is to be released soon. What does that mean to those of us with large twisted pair installations? Cheers, Ian
ecf_hap@jhunix.HCF.JHU.EDU (Andrew Poling) (10/04/89)
In article <42457@sgi.sgi.com> ian@lassen.wpd.sgi.com (Ian Clements) writes: [...] > I disagree. Remember that you need one transceiver for each workstation >regardless of whether or not you use thick, thin or twisted pair (assuming that >no workstation already has a thinnet xcvr installed). I'm calling the xcvr >costs a wash even though there is a $50 difference between thin and twisted >pair (thin being least expensive and twisted pair most expensive). I've been watching this discussion go by and I want on. Over and over I've been muttering to myself, "But they've probably already bought a thinnet xcvr - whether they know it or not". Fact is, a whole bunch of workstations come from the factory with a xcvr on board and that cute little BNC connector on the back. And what about PCs/compatibles? Almost EVERY ethernet card that I've laid eyes on for PCs has both thick and thinnet interfaces. Some machines are even available ONLY with thinnet interfaces (DEC VS2000's come immediately to mind - there must be others, though). In order to put a machine like this on thicknet or twisted-pair, you must insert a $1,000-or-so repeater of some sort. Uh-oh, kinda makes those sweeping generalizations start looking like expensive oversights. When considering the relative costs of thicknet, thinnet, and twisted-pair, you have to consider that alot of machines come thinnet ready. If we're figuring the costs of putting such machinery on alternate cable types, we have to admit that we're buying two xcvrs for each machine and leaving one idle. We may also end up jacking up our repeater-count converting from one cable type to another. We've been using thinnet more and more, recently, for in-building wiring for several reasons: 1) it's smaller in diameter and more flexible and thus easier to put in place than thicknet (twisted-pair likewise) 2) thinnet cable termination is a breeze compared to xcvr cable termination (we're talking about putting a connector on the end of the cable - fabricating custom-length cables) 3) when we want to put several machines in one room, we can daisy-chain 4) people with thinnet-ready equipment save money - no additional cost for a xcvr 5) people with DB-15 interfaces spend less because thinnet costs considerably less than xcvr cable per foot and we can put the thinnet xcvr very close to the machine We still use thicknet for the "backbone" (inter-building cabling) and for some "building risers" for mostly obvious reasons: 1) greater allowable cable length - important with our geography and layout 2) greater (perceived at least) durability under the sometimes adverse conditions encountered - I'm not sure that there isn't sufficiently sturdy thinnet cable available nowadays; in fact, I'm sure there probably is 3) the best reason of all - alot of our cabling of this type predates the wide availability and usage of thinnet I mostly wanted to point out some easily overlooked "hidden" costs. It's amaizing how fast the costs can climb when you start figuring in repeaters and pairs of xcvrs. -Andy -- Andy Poling Internet: andy@gollum.hcf.jhu.edu Network Services Group Bitnet: ANDY@JHUVMS Homewood Academic Computing Voice: (301)338-8096 Johns Hopkins University UUCP: mimsy!aplcen!jhunix!gollum!andy
roy@phri.UUCP (Roy Smith) (10/04/89)
In <2781@jhunix.HCF.JHU.EDU> andy@gollum.hcf.jhu.edu (Andy Poling) writes: > We still use thicknet for the "backbone" (inter-building cabling) and for > some "building risers" for [...] greater (perceived at least) durability > under the sometimes adverse conditions encountered I suppose the thick trunk cable is pretty tough, and the way the typical vampire tap xciever connects to the cable is pretty good, but the stupid D-15 connectors for the xciever drop cable is a disaster. The xciever ends don't give us any trouble because they are hidden away in a ceiling or wall where nobody can touch them, and we have the drop cable firmly lashed to the trunk cable with 2 or 3 nylon cable ties to keep them from shifting. On the other hand, the connection from the xciever cable to the workstations is our single most common cause of network failures. You just can't take a stiff heavy cable and expect it to stay attached to a flimsy slide-lock widget, especialy where the cable and/or the workstation are capable of being moved by accident. The Sun 3/50 is about the worst in this respect. We've totally given up on thick cable for PC's too. Our DEC, Kinetics, and TCL gear seem to have somewhat better designed slides, but still suffer from the inherent absurdity of the basic design. Our UB gear has the slide locks replaced with screws. It makes for a non-standard cable, but at least it doesn't fall out (this is, by the way, about the only good thing I can think to say about our UB ethernet bridges). We have a few machines in one office suite on thin (with a DEC DESPR thick-to-thin repeater) and have never had any trouble with the connections at all. -- Roy Smith, Public Health Research Institute 455 First Avenue, New York, NY 10016 {att,philabs,cmcl2,rutgers,hombre}!phri!roy -or- roy@alanine.phri.nyu.edu "The connector is the network"
jfinke@itsgw.rpi.edu (Jon Finke) (10/04/89)
In <4028@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: > I suppose the thick trunk cable is pretty tough, and the way the >typical vampire tap xciever connects to the cable is pretty good, but the >stupid D-15 connectors for the xciever drop cable is a disaster. The >xciever ends don't give us any trouble because they are hidden away in a >ceiling or wall where nobody can touch them, and we have the drop cable >firmly lashed to the trunk cable with 2 or 3 nylon cable ties to keep them I have had cables like this cause problems when someone else pulls a wire throught the ceiling, or opens the ceiling for any other reason. >from shifting. On the other hand, the connection from the xciever cable to >the workstations is our single most common cause of network failures. You >just can't take a stiff heavy cable and expect it to stay attached to a >flimsy slide-lock widget, especialy where the cable and/or the workstation >are capable of being moved by accident. We also found this to be our most common ethernet failure at RPI. It is now standard practice here to replace slide lock hardware with screw lock hardware for all installations. We mostly use thinnet, but enough equipment comes through with the DB15s that we still convert when it arrives. About half the equipment can be modified without opening the covers. I don't think we have had a failure of a screw lock connected DB15. We also make a lot of our own drop cables, that way we can get the correct hoods and hardware for them. Some cables can also be modified for screwlock hardware. This does preclude the use of right angle cables, but that seems a small price to pay for the greatly increased reliability. -- Jon Finke jfinke@itsgw.rpi.edu Network Systems Engineer USERB239@RPITSMTS.BITNET Information Technology Services 518 276 8185 (voice) Rensselaer Polytechnic Institute 518 276 2809 (fax)
henry@utzoo.uucp (Henry Spencer) (10/05/89)
In article <4028@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: > The Sun 3/50 is about the worst in this respect... Our DEC, Kinetics, > and TCL gear seem to have somewhat better designed slides... It's worth pointing out that the problems on the Suns are not just poor design, they are out-and-out violations of the specs for the connector mounting. The connectors never get a chance to seat to their full depth. Actually meeting the letter of the law of the specs takes work and is a bit inconvenient; Sun didn't bother. -- Nature is blind; Man is merely | Henry Spencer at U of Toronto Zoology shortsighted (and improving). | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
earle@poseur.JPL.NASA.GOV (Greg Earle - Sun JPL on-site Software Support) (10/08/89)
In article <1989Oct4.153141.19593@rpi.edu> jfinke@itsgw.rpi.edu (Jon Finke) writes: >It is now standard practice here to replace slide lock hardware >with screw lock hardware for all installations. We mostly use >thinnet, but enough equipment comes through with the DB15s >that we still convert when it arrives. Please, people, think about what you are talking about before you post. There is no reason a thread about Ethernet cabling and DB-15 connectors should be cluttering up the TCP-IP mailing list (a.k.a. comp.protocols.tcp-ip). Please take it over to comp.dcom.lans where it belongs. (We now return you to the "ASCII vs. PostScript" 15-round title bout (^: ) -- Greg Earle Sun Microsystems, Inc. - JPL on-site Software Support earle@poseur.JPL.NASA.GOV (direct) earle@Sun.COM (indirect)