jqj@svax.cs.cornell.edu (J Q Johnson) (08/02/86)
I'm planning Ethernet wiring for a new building (about 150 stations). I'd been planning a traditional scheme with thick backbone+xcver in hallway raceways and drops into each office, similar to my existing offices. However, I am investigating using thin Ethernet instead. I'd appreciate advice from one of you who has installed thin Ethernet in a multiple-individual-office environment. The environment is quite multivendor, with a few VAXes, 50 Xerox Dandelions, perhaps a dozen each of SUNs and Symbolics Lisp Machines, and lots of other vendors' hardware. Also, most offices currently contain only a single workstation, though this may change in the next few years. A thin ether configuration would presumably consist of a DEC DEMPR or equivalent, with one RG58 leg running down each hallway, looping into each office (say 8 offices) for a 300' maximum length. In each office the RG58 loop would appear at one point on a wallplate. This might be the base of a T-connector (allowing a cheapernet xcver like the DEC DESTA to be attached directly to the wall) or a BNC connector with a small external loop of RG58, allowing extension of the RG58 into the office. Does this latter scheme introduce too many connectors and hence reflections on the ether? Does having connectors in each office make it likely that users will accidentally disconnect the Ethernet cable? Are there better plans? I'm also interested specifically in the DEC thin-ether xcver (DESTA). Does anyone know how compatible it is with non-DEC interfaces? Does it provide (disableable) heartbeat like the H4005? Are there preferable alternatives from other vendors? Can I hook a DELNI to a DESTA? I don't completely understand the configuration issues for thin ether. RG58 of course has greater attenuation and slower propogation than does thick ether. What is the delay budget, etc. for a multiport repeater like a DEMPR (is it just like a standard repeater)? DEC's "networks" guide states that a thick segment connected to a DEMPR must be less than standard length and have no other repeaters; why? Are lots of BNC connectors in the thin segments a problem? How about transitions between thick and thin ether cable (would I do well to replace long straight runs of RG58 with transitions and thick ether)? Anything else I should beware of? Probably best to reply by mail unless you think your answers are of general interest: jqj@systems.cs.cornell.edu [preferred] jqj@cornell.arpa OR jqj@crnlcs.bitnet OR jqj@cornell.uucp
dennisg@pwcs.UUCP (Dennis Grittner) (08/07/86)
In article <436@svax.cs.cornell.edu> jqj@systems.cs.cornell.edu (J Q Johnson) writes: >I'm planning Ethernet wiring for a new building (about 150 stations). >I'd been planning a traditional scheme with thick backbone+xcver in >hallway raceways and drops into each office, similar to my existing >offices. However, I am investigating using thin Ethernet instead. > I would like responses on this to be posted to the net. I have similar questions about thin ethernet and mixing thick and think as well as the specific devices for doing this AND is any/all of this a good idea. Thanks in advance.. Dennis Grittner Public Works Computer Services 612-298-4402 Room 700, 25 W. 4th St. St. Paul, Minn. 55102 I must have been asleep for twenty years - - Ronald ( Bonzo ) Reagan is president? - - Rip von Grittner - 1986
michaels@hplabsb.UUCP (Robert Michaels) (08/12/86)
In article <210@pwcs.UUCP>, dennisg@pwcs.UUCP (Dennis Grittner) writes: > In article <436@svax.cs.cornell.edu> jqj@systems.cs.cornell.edu (J Q Johnson) writes: > >I'm planning Ethernet wiring for a new building (about 150 stations). > >I'd been planning a traditional scheme with thick backbone+xcver in > >hallway raceways and drops into each office, similar to my existing > >offices. However, I am investigating using thin Ethernet instead. > > > I would like responses on this to be posted to the net. I have > similar questions about thin ethernet and mixing thick and think > as well as the specific devices for doing this AND is any/all of > this a good idea. I've installed a so called thin ethernet here in 150+ workstation environment. All the workstations are HP 9000s running Unix. In addition there are several mainframes connected to this same network. Thin ethernet was helpful for us because we have an open office environment which uses cubicles instead of walls. We just drop the cable from office to office, and make sure there is a BNC "T" connector in each office. Although the thin segment can be up to 185meters and 30 connections I try to keep the length and number to less than 100m and 10 connections. This way I'm flexible for expansion or change. I use a multi-port repeater built by HP called the HP 28645A which connects 4 thins to single thick cable. It has half the delay of a standard repeater which means that if you had no other repeaters on the net you could cascade these guys 4 deep. Although I can't say for sure unless DEC really screwed up their multi-port repeater should have similar characteristics. What I have found in running this thin stuff is that buying preconnected cable is a major win. Also, trying to put a wall plate in each office is not worth the effort. I figure how much cable it takes to go from office to office and buy a bunch of cables for that length. Then each office essentially has two cables coming in connected by a T. If someone decides to add another host(s) in that office I just insert as much cable and Ts as is necessary. BTW: There is no minimum distance spec necessary to maintain in thin LAN (ie: you can place connections as close together as you want). The multi-port repeaters connect to a backbone(s) of thick cable which runs between the floors of this building and finally ends in the computer room where it ties into the mainframes. One advantage of the multi-port repeaters is that they provide electrical isolation to the remainder of the network. If any individual thin segement gets unterminated or damaged for some reason the repeater isolates that segment. If you need to change a particular part of the network you don't run the risk of bothering all the users on the net by making a bad tap. Besides, this RG-58 stuff is a lot easier to pull than a long tranceiver cable or the yellow cable. Hope this helps. I can tell you about how we are using multiple thick backbones, ethernet patch panels and IP gateways in our environment if you are interested. - Robert ( michaels@hplabs ) DISCLAIMER: Please understand that these are my opinions and do not necessarily state official HP policy.
ray@ssl-macc.co.uk (Ray Saxton) (08/21/86)
We have a site with 16 suns and 2 vaxes, and have been running thinnet with 3Com boxes for about a year now with no major problems. Each office has several "barrel" connectors available for hooking in transceivers as required. Although 3 floors of the building were wired as a single segment of some 200 mtrs, it has been found prudent to trim this as transfers between the end nodes were not 100% reliable. We have therefore bypassed , or just plain ignored, some 30% of the net and now have no problems. I say that we have no problems now , but starting up was a pain. Firstly loose connections proliferated, despite having professionals in to do the job. As suggested preformed cables would cure this one . Secondly and more annoyingly was the reaction of both DEC and SUN "engineers" on seeing the thinnet. That won't work. Thats not 10mtrs ....on and on and on they went.We still have the state where VAX A runs a H4000 into the net and works, Vax B runs an H4000 into the net and works to some sites . This is blamed on the thinnet by both sets of Engineers, despite tests that proved that cable length did not affect the ability of certain sites to communicate nor the inability of the others to communicate . We have finally resolved the problem by hooking the VAxes via Delni and then hooking that into the despised thinnet via the good H4000. My only worry is what happens if that H4000 goes to meet its forefathers. -- Ray Saxton ray@uk.co.ssl-macc UK PSS 2342 672500104 mcvax!ukc!dcl-cs!sslvax!ray Telephone +(44) 625 29241
phil@amdcad.UUCP (Phil Ngai) (08/26/86)
In article <362@ssl-macc.co.uk> ray@ssl-macc.co.uk (Ray Saxton) writes: > > We have a site with 16 suns and 2 vaxes, and have been running thinnet >with 3Com boxes for about a year now with no major problems. Each office >has several "barrel" connectors available for hooking in transceivers as >required. Although 3 floors of the building were wired as a single >segment of some 200 mtrs, it has been found prudent to trim this as >transfers between the end nodes were not 100% reliable. According to the DEC _Networks and Communications Buyer's Guide_, page 2.102, "ThinWire Ethernet is good for 185 meters and only 30 devices. How many devices did you have? >and more annoyingly was the reaction of both DEC and SUN "engineers" >on seeing the thinnet. That won't work. Thats not 10mtrs ....on and on and on What? >This is blamed on the thinnet by both sets of Engineers, despite tests >that proved that cable length did not affect the ability of certain sites >to communicate nor the inability of the others to communicate. I wouldn't consider this a definitive test. Some devices may have more margin than others, that doesn't mean your cable plant isn't beyond spec. In particular I believe that 3Com boxes are designed with extra margin and so their ability to work doesn't mean you can expect other Ethernet compatible equipment to work. > We have >finally resolved the problem by hooking the VAxes via Delni and then hooking >that into the despised thinnet via the good H4000. My only worry is >what happens if that H4000 goes to meet its forefathers. Sounds like you're living on borrowed time to me. -- Rain follows the plow. Phil Ngai +1 408 749 5720 UUCP: {ucbvax,decwrl,ihnp4,allegra}!amdcad!phil ARPA: amdcad!phil@decwrl.dec.com