Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!cs.utexas.edu!uunet!usna!tsmith From: tsmith@usna.MIL (Tim G. Smith ) Newsgroups: comp.dcom.lans Subject: Re: Simple thin ethernet question Message-ID: <194@usna.MIL> Date: 28 Jul 89 17:07:18 GMT References: <2946@ukecc.engr.uky.edu> <383@belltec.UUCP> Reply-To: tsmith@cad.usna.mil.UUCP (Tim G. Smith (Mechanical) ) Distribution: usa Organization: U.S. Naval Academy, Annapolis Lines: 112 > 2) You have picked a bad time to ask advice from the net. > If you run thin Ethernet, aka rg-58, through the wall, you'll be > sorry. Morning, noon and night. You'll go to the Justice > of the Peace and change your name to Job. Thin Ethernet is > unreliable (in particular the connectors) and was designed > as a cheap alternative to Ethernet. It is only suitable for > connecting one or more computers in the same room. > >Ethernet through the wall should today be thick Ethernet (big yellow cable) >or tranceiver cables. You should put a 4-, 8-, or 16-port multiport >transceiver in each room which has more than one computer. > >Ethernet through the wall tomorrow should probably be done with the >Ethernet-over-Twisty-Pair standard that is working it's way out of the womb. >(Ignore all recommendations about Twisty Ethernet until it's a standard, >and standard equipment has been shipping for at least 6 months so that the >industry can discover the gotchas by living with them.) > >Lance Norskog >Sales Engineer >Streamlined Networks Ummm exscuse me, but can you please explain your reasons for making statements likes the above. Please provide case studies to back up your complaints against thin ethernet. I have the exact opposite feelings. Thick ethernet is a royal pain and can be a major nightmare to deal with. At my former job I personally removed almost all of the thick ethernet cable and replaced it with thin. The network in question consisted of at least 5 subnets and a whole lot of cable. A poorly desinged thin net can be hellish but not much worse then the average thick net. I designed my thin net as a star. A single RG58 was run to each office/lab with all of the wires for each floor starting in a central closet where there was a rack with Multi Port Repeaters (Cabletron MR900Cs and Cabletron MMACs). From each floor at lest one (and ususally more) RG58 was run down to the main computer room where the network head was located (the backbone linking all of the various thin/thick/FO subnets was completely located in one rack). The cost of the additional cable is trivial- the cost of the MPRs and/or MMACs is not too outrageous for the convenience, flexibility, reliability, and sanity they provide. Some of the advantages of star wiring are: 1) Reconfiguring the network is trivial. If office X changes hands from being in the MechE dept to EE then it takes all of two minutes to move the office from the MechE subnet to the EE subnet. Try doing that with any sort of daisy chained wiring plan- not easy. 2) MPRs (or MMACs or any other fully compliant IEEE 802.3 repeater) will isolate any faulted segment from the rest of the network. So if someone decides to take the net apart in his office the only folks affected are the guilty party and the guilty party's office mate. 3) Locating problems with a star network is much easier than with a daisy chain scheme. You simply go look at the MPRs and see if there are any fault lights on. If so you know that the problem is between the office and the MPR so you just fire up the TDR and see what is going on (this assumes that when the wiring was installed it was TDRed from both ends and the measurments recorded). If the problem is not between the office and the MPR than it must be in one of the other segments leading back to the network head (provided there is one)- if the network has been designed well there should be cleary defined segments that can be debugged one by one until the problem is found. Some other reasons I like thin net: - no vampire taps! - connectorizing is trivial- anyone can do it with a bit of practice - the wire is smaller and easier to work with - many hosts have both BNC and DIX plugs on them- if you use the BNC you do not need an external XCVR - neat repeaters like the Cabletron MMAC exist that can collect very useful network loading statistics on each cable segment and can also support redundant links between the MMACs which means that your most critical segments can be well protected from failure. With all of that said the most important thing is to design the network carefully. A well designed network can use any media and be reliable- I just happen to prefer thin net as my media. Twisted pair also is pretty neat, I just don't have much experience with it. Star wiring and network heads with "backbones in a rack" are real winners when it is debug time even if they cost more to implement. Having the right tools to debug is also essential. I would not build any network without having a TDR for each type of media I use (ok- I have to admit I did use fiber without haing a FO TDR but I did have some other stuff including a LED attached to a 9 volt battery for shining down the fiber and I did eventually get a signal meter from Siecor for about $800). There is nothing wrong with thin net- just with poorly designed thin nets. later, Tim Smith (formerly of the US Naval Academy, and still getting news at usna.) US mail:US Army, BRL E-mail: SLCBR-SE internet:tsmith@brl.mil Aberdeen Proving Grounds, MD 21005-5066 uucp :...!uunet!brl!tsmith MaBell :(301)278-6678 (or 6808) Autovon: 298-6678 Disclaimer: The DoD does not pay me enough to speak for them.