aiko@cs.odu.edu (John K Hayes) (10/30/90)
Anybody know the advantages and disadvantages of ethernet and arcnet? I know ethernet is faster and seems to be cheaper as well. What are the problems involved with a single workstation screwing up the whole net with ethernet as opposed to arcnet? I heard from someplace that with ethernet, if a station goes down (locks up), it affects the rest of the net (possibly brings the whole net down) - but I don't think this is true. I heard also that it is possible for one workstation to mess things up (not necessarily bring down) for the whole net with both ethernet and arcnet. How's that? I have seen with my own eyes where a arcnet setup was brought down by a faulty cable connection to one of the workstations (this is reasonable); I'm interested in software related problems (such as lock ups). I am setting up a lan for 6-10 PCs with a possible expansion to 20-50 PCs each no more than 500' apart. Any comments on the type cabling I should use? I am shying away from twisted pair since it seems to be cost prohibitive in terms of network adaptor cards and concentrators and also worries me about line noise and unshieldedness. On the other hand, I am leaning toward unshielded coax since it is better than unshielded twisted pair and shielded coax is too expensive. Do you think this will be sufficient? Also - can anyone explain (generally) the technology behind ethernet and arcnet? As in how they work - why one is faster than the other - which one is best suited to a net with less than 50 PCs. Thanks loads, John Hayes -- ---{john hayes} Old Dominion University; Norfolk, Virginia USA internet: aiko@cs.odu.edu Work: (804) 460-2241 ext 134
hedrick@athos.rutgers.edu (Charles Hedrick) (10/30/90)
Ethernet and arcnet are similar technologies, except that Ethernet is faster. It's less expensive because there are much more Ethernet users, so production volumes are higher. If a station goes down on Ethernet, there is no effect. It is in principle possible to shut down any network due to one station doing something wrong, if the things is sufficiently pathological, but the Ethernet design attempted to protect against as many as possible. Actual results are that they were pretty successful. If a station just has its software hang, that's not even a dangerous situation for the network. Of course if the software gets into a loop sending packets continuously, it's pretty hard to expect any network technology to do much to fix that. It won't necessarily take down the network, but it will certainly create a heavy load. Thinnet (thin coax) is generally cheapest for a lab situation, i.e. lots of machines in one room. The disadvantage is that it sort of like old Christmas tree lights (or Localtalk), in that one person disconnecting their cable will disrupt the network. If it's in one room this shouldn't be a problem, but with lots of user offices it could be. The reason thinnet is cheap is that it needs no electronics beyond what's on the controller card. Beyond that, the usual recommendation these days is unshielded twisted pair. It's easy to expand, and easy to do network diagnosis on. It also has a single point of failure, the box in the wiring closet that all the wires go to. This seems not to bother people. I sort of like the original thick coax myself. But I have to admit if somebody blows installation of a tap, it can short the cable and you may have a hard time finding it. The nice thing about the twisted pair stuff is that the wiring to one office can't affect other offices. In this respect it's better than the other Ethernet technologies (or Token Ring). Basically though any of the wiring schemes have their advantages and disadvantges, and probalby about all work out the same, as long as you are careful.
nelson@sun.soe.clarkson.edu (Russ Nelson) (10/31/90)
In article <Oct.29.23.23.03.1990.3907@athos.rutgers.edu> hedrick@athos.rutgers.edu (Charles Hedrick) writes:
Of course if the software gets into a loop sending packets
continuously, it's pretty hard to expect any network technology to
do much to fix that. It won't necessarily take down the network,
but it will certainly create a heavy load.
Ha! I was working on a little Ethernet monitoring program (using
Clarkson's packet drivers, of course) that put the packet driver into
promiscuous mode. That's all well and good, but I forgot to take it
*out* of promiscuous mode. Then I ran Phil Karn's TCP/IP package,
which likes to do IP routing. It tried to route *every* packet
on the net back out onto the same net!
This put a heavy load on the net, but it also greatly confused our
sun workstations. It was kind of neat. I could unplug my machine,
and the net worked again. Plug it back in again, and viola, instant
confusion.
--
--russ (nelson@clutx [.bitnet | .clarkson.edu]) FAX 315-268-7600
It's better to get mugged than to live a life of fear -- Freeman Dyson
I joined the League for Programming Freedom, and I hope you'll join too.
lars@spectrum.CMC.COM (Lars Poulsen) (10/31/90)
In article <Oct.29.23.23.03.1990.3907@athos.rutgers.edu> hedrick@athos.rutgers.edu (Charles Hedrick) writes: > ... The nice thing about the twisted pair stuff is that the wiring to >one office can't affect other offices. In this respect it's better >than the other Ethernet technologies (or Token Ring). It is perfectly legal to wire thin ethernet with a repeater in the wiring closet and a coax to each station. When you do that, you gain the same advantage. At about the same cost. If you are pulling new cable for each workstation, the costs for thin ether and twisted pairs are comparable (except that you get up tp three pairs for the price of one). If you are consulted before the building goes up, the same applies. The best thing about twisted pairs is that sometimes the builder put in a few extra, so there is cable available without any wire pulling. -- / Lars Poulsen, SMTS Software Engineer CMC Rockwell lars@CMC.COM