Article View: vmsnet.networks.misc
Article #18Re: Q: Extending Ethernet over fiber
From: Lambert Hofstra
Date: Mon, 16 Nov 1992 23:03
Date: Mon, 16 Nov 1992 23:03
112 lines
6854 bytes
6854 bytes
In <1992Nov12.074211.3898@wkuvx1.bitnet> goathunter@wkuvx1.bitnet writes: >We have a VAX 6320 with a small Pathworks LAN running over Thinwire >Ethernet; this is in one building. The 6320 and two DECserver 500s >are connected to a DELNI that's connected to a DESPR. I plan to >eventually remove the DELNI from that configuration, I think. We'd >like to extend our Ethernet in two different ways as cheaply as >possible (of course). >I have two configuration questions for those of you who know about >Ethernet. > o The building housing the above LAN is connected to another > building via fiber optic cable, but it's not being used yet. > We have a Novell LAN in that building, as well as a DECsystem > 5100 and a VAX 11/785. We'd like to extend the 6320 LAN to > include at least the DECsystem and the 11/785, with the Novell > LAN being a nice added bonus. > The question: if we buy two Fiber Media Access Units (DEFLM-AA) > as advertised on page 105 in the new DECdirect catalog, can we > connect the 6320 and the remote systems without purchasing any > additional equipment (routers, bridges, etc.)? If so, do you > have any suggestions for configuration? The remote building is > about 500 feet away; the Pathworks LAN is currently about at its > length limit. Approximately 200--300 feet run between the > DECsystem and the 11/785; the DECsystem is currently connected > to the Novell LAN. > It appears that the DEFLM would let us go from Ethernet -> DEFLM > -> fiber -> DEFLM -> Ethernet. > o We have another Novell LAN (Thinwire Ethernet) in the same building > as the 6320. It too has reached its maximum length. Naturally, > we'd like to connect the Novell LAN and the Pathworks LAN. Will > that work if the DESPR is replaced with a DEMPR? How about just > connecting a gateway to the DELNI now? Would that work? >Thanks in advance for any help you can provide. No one around here >(including our DEC reps) has had much experience with the limits of >Ethernet. >Hunter You will have to look at two kinds of "maximum length", the first is the maximum lenght of the cable without loosing too much signal strength, about 100 meters for twisted pair, 180 for thinwire and 500 meters for the big yellow coax cable. These pieces of cable can be extended with a repeater, that amplifies the signal. The second maximum length is not measured in "real" distance, but in microseconds. Since ethernet is a broadcast type of network, inwich collisions can occur, it is necessary to be able to detect these collisions. This means that the transmitter must check the cable to make sure nothing went wrong. It must check until the signal can travel from the transmitter to the other side of the network and back. Because of this, ethernet distances are measured in microseconds. Every unit that is used has its specific propagation delay, and thus adds to the total length, although it might only have a physical length of 0.05 meters. The big problem is how to find these propagation delays, they're not in the installation guide. Every repeater you add, adds a rather large propagation delay, and thus a "large distance". DEC advices not to use more than two repeaters between any two systems (VAX, PC's, bridges) on the network. (A bridge does complete retiming on packets, so lenght calculations start all over if a bridge is passed). This does not mean that you can not have more than two repeaters as a total: a backbone can have several repeaters, as long as you don't cascade them, the total number of repeaters betwqeen systems will be two! Now in your case, let's look at the possibilities: Your "backbone will be the fiber between the building, so you would have to use a bridge, or have a maximum of one repeater in each building, for instance a DEMPR at the side of the 6320. You can connect the two thinwire segments, and would have to connect the 6320 with a DESTA. Don't know if the DECserver 500 has a thinwire, if not, you would have to use DESTA's for them too. (both DELNI and DEMPR/DESPR are repeaters!) At the other building you might be able to use a DEMPR to connect the NOVELL segment, and to connect the 11/785 with a DESTA on thinwire. Now this might work, but given the maximum length of the cables in use, i would not make a bet on it. My advise would be to use a bridge in between, for instance the DECbridge 90, which is rather cheap. Any other will do, given the right connections. This way you would have to use a DEMPR on the 6320 side if you want to add the NOVELL network, otherwise you can still use the DESPR, and connect the AUI output of the DELNI to the Fiber optic tranciever. At the other side, connect the Fiber optic tranciever to the bridge, and connect the other side to the network overthere, for instance with a DEMPR connected to the NOVELL part, and another thinwire segment to the 11/785 (with a DESTA). By the way, one of the ways to find out if your ethernet network is to long, is to boot a terminal server at one side, and see if you get error messages (on a VAX with DECnet) like "remote node unable to defer, distance 15384". The distance in meters can be calculated if you know the delay of the controllers, the tranceivers, the different cable types, the repeaters etc. This message will show up more often on a busy network, not if all systems are down. The message means that (at least) two systems started to transmit, and only one has seen that a collision has occurred, since the collision was to far away for the second system: the collision was detected after the second system had concluded that everything was send correctly. Booting terminal servers in such a network will almost always work, but somethimes will result in a communication error. Hope this helps a little. If someone has a list of propagation delays for different types of ethernet equiptment, please send it. >------ >Hunter Goatley, VMS Systems Programmer, Western Kentucky University >goathunter@WKUVX1.BITNET, 502-745-5251 Lambert Hofstra lamhof@thijssen.nl
Technical Details
Message-ID:
<1992Nov16.230355.6959@thijssen.nl>Path:
arc1-ca.pugleaf.net!sparky!uunet!know!mips2!news.bbn.com!noc.near.net!news.Brown.EDU!qt.cs.utexas.edu!cs.utexas.edu!sun-barr!west.West.Sun.COM!grapevine.EBay.Sun.COM!news2me.EBay.Sun.COM!cronkite.Central.Sun.COM!texsun!exucom.exu.ericsson.se!ericom!sunic!mcsun!sun4nl!troy!lamhofReferences:
<1992Nov12.074211.3898@wkuvx1.bitnet>Show Full Headers
Complete Article Headers
Path: arc1-ca.pugleaf.net!sparky!uunet!know!mips2!news.bbn.com!noc.near.net!news.Brown.EDU!qt.cs.utexas.edu!cs.utexas.edu!sun-barr!west.West.Sun.COM!grapevine.EBay.Sun.COM!news2me.EBay.Sun.COM!cronkite.Central.Sun.COM!texsun!exucom.exu.ericsson.se!ericom!sunic!mcsun!sun4nl!troy!lamhof From: lamhof@thijssen.nl (Lambert Hofstra) Newsgroups: vmsnet.networks.misc,vmsnet.misc,comp.os.vms Subject: Re: Q: Extending Ethernet over fiber Message-ID: <1992Nov16.230355.6959@thijssen.nl> Date: 16 Nov 92 23:03:55 GMT References: <1992Nov12.074211.3898@wkuvx1.bitnet> Organization: Thijssen BV, Veenendaal, The Netherlands Lines: 112