Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!spool.mu.edu!uwm.edu!ux1.cso.uiuc.edu!uxh.cso.uiuc.edu!german From: german@uxh.cso.uiuc.edu (Gregory German) Newsgroups: comp.dcom.lans Subject: Re: 10Base-T hubs Message-ID: <1991Apr10.150801.2519@ux1.cso.uiuc.edu> Date: 10 Apr 91 15:08:01 GMT Article-I.D.: ux1.1991Apr10.150801.2519 References: <1582@vtserf.cc.vt.edu> <1991Apr9.181721.15560@leland.Stanford.EDU> <1991Apr10.035143.15862@leland.Stanford.EDU> Sender: usenet@ux1.cso.uiuc.edu (News) Distribution: usa Organization: University of Illinois at Urbana Lines: 104 morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes: >> Perhaps I'm being too stupid here, but why do you need a repeater >> for an 8-station network? Why not hook them all up to the same >> cable? >Well, if you've been reading comp.dcom.lans, the benefits of >structured wiring have been explained at length. But this is indeed >just the sort of question that those of us who install these things >for a living have to answer on a daily basis. You're right, it isn't >obvious. So here are the answers in easy-to-understand terms. >1. I didn't say so in my previous note, but the main point of putting >in a network around here (and at almost any large institution these >days, I'll bet) isn't to hook the machines to themselves, it's to hook >them to the campus net (that is, to the Internet). So your proposed >single run of thinnet has to connect to the building backbone in any >case, which means a repeater. I'll admit I can't explain it, but I >can get a 12-port 10Base-T repeater these days for about the same >price as a traditional 2-port AUI-only repeater. >2. Where are these 8 stations anyway? As I said before, if they're >in one room, sure, use thinnet. But if, as is the usual case, they're >scattered around the floor of your medium-to-large building, then the >"single cable" ends up hopping from room to room, crossing the hall, >going around the corner, and doubling back to catch the one at the end >of the hall. Maybe you're lucky and it's still only 450 feet. >Then they get two more computers in two different offices (and yes, >they *always* get more computers eventually). Someone looks up in the >ceiling and finds where the cable runs kind of near the one office, >cuts it and adds a loop. The other one is sort of near the end of the >daisy-chain, so they just add to the end. How long is the cable now? >Where are its connectors that will fail over time? Who bought the new >cable that was spliced in, and why didn't they notice that it was 53 >ohms, not 50 (they're lucky it wasn't 75)? If the UTP wire already pulled is useable then it is a simple matter to "pre-wire" every phone location. If you use a slotted concentrator you only have to add enough ports to meet your current needs and adding ports as needed. This is a big win over either pulling coax to EVERY location where you might want a computer or having a contractor come in and add cable on a case by case basis which will end up costing you more per connection than doing it all at once. >So eventually the thing breaks down, and after weeks of cursing and >head-scratching, they call in the network guys, who look around, shake >their heads, and say: what you should do is buy a MPR and install some >home runs to it. That way you'll have a manageable cable plant that >is relatively immune to the weird things that people do to nets, and >that provides an obvious growth path for future installers. I concur with Bob on this. The more networks you install, the more you will realize that the cost of doing it right outweighs the cost of the time lost. While it is true that the end users will rarely (if ever) give you a second thought if your well designed network does its job there is no quesiton who they will blame if the network is down because of cable problems. >3. If those 8 stations are typical office PCs these days, they're >probably worth a total of $20,000 or so. If they're fancy Unix boxes, >the total may be over $100,000. If the people using them are getting >paid for their time, the total may be a couple of hundred dollars an >hour (of course, if they're just grad students ... 8^). Isn't it >worth spending a extra few hundred dollars, adding at most 3 or 4 >percent to the total cost of this networked computer system, to make >the network as reliable as it can possibly be? >So, the point is: no matter whether you're installing thinnet or >twisted-pair or fishing line, it's almost always worth it to install >a structured wiring plant. And once you've decided to do that, *if* >you can use your existing structured telephone cable plant, then >10Base-T is cheaper by a long shot. No hype. In those cases where I do not have a useable UTP cable plant for 10baseT I still use thinnet, but in most every case every office/lab will have its own port on a multi-port repeater. Flexibility and reliability are my prime motivating factors. Each office is on their own home run and somewhat isolated from the "usual" problems of the daisy-chained environment and switching them from one network to another is as simple as choosing which repeater they are connected to in the closet. UTP is even more flexible since I can pre-wire 2 pair from each office to a patch pannel and then use that wire for 10baseT, TRN, AppleTalk, 3270 or serial connetions. I am finding out that I have users on Macs that start out on AppleTalk and then upgrade to ethernet. All I have to do in most of my networks is remove the jumper cable between the StarController and the patch pannel associated with that office and replace it with one to a 10baseT hub. Even though we don't use it in most cases I have also designed in the ability to add intelligent monitoring tools to the the hubs. LANview with Cabletron repeaters and an module with SNMP capability in the Plexcom chasis. I expect to see increased use of these as tools become available. > - RL "Bob" Morgan > Networking Systems > Stanford -- Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) US Mail: Univ of Illinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801 Office: 129 Digital Computer Lab., Network Design Office