bronson@mfci.UUCP (Tan Bronson) (03/25/89)
Newsgroups: comp.dcom.lans
Subject: Thin ethernet question
Expires:
References:
Sender:
Reply-To: bronson@mfci.UUCP ()
Followup-To:
Distribution:
Organization: Multiflow Computer Inc., Branford Ct. 06405
Keywords:
We're getting ready to wire our new building for ethernet.
We'd like to have five physical networks (for each group of workstations),
these networks will be linked to a backbone using Retix 2244 bridges. The
backbone will exist in the computer room only.
We have closed offices in this building, so we want one or two
ethernet taps per office. It's a new building, so adding the wiring
is easy so long as I decide in the next week or two.
I've looked into several options:
1) Using twisted pair ethernet. I looked at Cabletron's system as an
example, but the cost to maintain multiple networks was too high.
The cost for the twisted<->thinethernet converter was $450,
and we needed one per office...
2) Having multiple yellow cable backbones thoughout the building.
This involves using a Delni (or a thin ethernet repeater) port
for each office. Thin ethernet is nice because a single connection
in a office can be used for more than one workrstation, but
the cost of a thin ethnernet repeater is $2300 versus $850 for
a Delni. This scheme relies on the locallity of people who should
be on the same network for efficiency, but people could move easily.
The cost of adding a user ranges from: adding another node on an existing
Delni if there is room, adding another tap on the appropriate yellow
backbone for a single user, and adding a tap and a Delni for two or
more people inthe same group.
3) Having multiple(many) thin ethernet `backbones'. The involves
wiring groups of 5-8 offices with one `backbone' and then running
this backone back to the wiring closet. The nice part is we
can have up to 8 thin backbones leading into a single thin ethernet
repeater, and we only need one thin ethernet repeater per network.
The bad parts are: 1) if someone moves into an office which has the
`wrong' backbone in it we're going to have to create another
backbone, and 2) we're going to have to get a TDR to ensure this
all works!
My boss likes 3) as it's a bit cheaper, and once done it should
not require going into the ceiling again. I like 2) using Delnis and
tranceiver cables, because it's easy, flexible, and I've has some bad
experience with thin ethernet. The major disadvantage of 2) is that
dual person offices require two tranceiver cables. (I'd prefer to do 2)
using thin ethernet cable, but the thin ethernet repeaters are too expensive!
I'm looking for experience and wisdom to help decide between
2) and 3) (or a better cost effective way of doing it.)
I also have a few questions about thin ethernet:
- How do other people wire large groups of workstations
using thin ethernet ?
- it looks like I need a 30' loop in each office so I can
connect a workstation 15' away from where the wire enters
the building, and I'd like to make the wiring neat! If I were
to put a panel with a pair of coax connectors in each office:
- would I be decreasing the number of nodes I can have ?
- I would effectively add a minimum of two segments to
add node onto the network. Is this fatal ?
thanks in advance
-------
Tan Bronson
Multiflow Computer Inc UUCP(work): {yale,uunet}!mfci!bronson
175 N Main St UUCP(home): {yale,mfci}!bronson!tan
Branford, Ct 06405 Phone(work):(203)-488-6090
--
Tan Bronson
Multiflow Computer Inc UUCP(work): {yale,uunet}!mfci!bronson
175 N Main St UUCP(home): {yale,mfci}!bronson!tan
Branford, Ct 06405 Phone(work):(203)-488-6090tsmith@usna.MIL (Tim G. Smith ) (03/28/89)
In article <724@m3.mfci.UUCP> bronson@mfci.UUCP (Tan Bronson) Asks about how to wire a new building for ethernet and writes: > 3) Having multiple(many) thin ethernet `backbones'. The involves > wiring groups of 5-8 offices with one `backbone' and then running > this backone back to the wiring closet. The nice part is we > can have up to 8 thin backbones leading into a single thin ethernet > repeater, and we only need one thin ethernet repeater per network. > The bad parts are: 1) if someone moves into an office which has the > `wrong' backbone in it we're going to have to create another > backbone, and 2) we're going to have to get a TDR to ensure this > all works! At my last job at the Naval Academy we spent a long time thinking about how to wire our buildings and eventually decided to use all thin ethernet and to use a star wiring configuration with each office wired back a the central wiring closet. Cabletron MR-9000 MPRs were used inside the wiring closet to to tie together all of the offices on one subnet. Thin coax runs were then used to run back to the central computer room where each subnet tied into the routers. If the distance from the wiring closet to the central computer room is over the limits (our typically weren't) point to point fiber systems can be used (Cabletron sells somes tuff that will do the trick- so does Codenoll). Once the data is in the computer room it can be tied together with bridges (blech! but if that is what turns you on go for it) or routers. All of the MPRs to do this were not cheap but in my opinion they were well worth the money. If you use MPRs you can save a lot of network debugging time- you just look for the faulted segment and then deal with it. The whole net does not come crashing done due to one fault. They are really great for making life easier on the network managers. They also keep one person from screwing up the net by pulling the network apart or taking of his terminator. If someone screws with the parts of the net accessible in his office he only takes himslf and maybe his officemate off the net. No one else gets affected. That is a real nice feature when you have a lot of naive users. Wiring each office individually also makes it easier when people move from one subnet to another (USNA subnets by departments and every summer offices get shuffled among departments). If all of the cables go back to the wiring closet you just switch the cable from one MPR to another when an office moves to a new subnet. (ie RM312 just became an Aerospace Eng office when it used to be a Mechanical Eng office- no problem move cable number 312 from the Aero MPR to them Mechanical MPR and everything is fixed). Another reason to use thinnet is that more and more machines are coming with on board XCVRs (and thus have bnc plugs on the back). All you do is run a piece of coax from the wall to the back of the workstation. If there is more than one machine in the office than you can daisy chain them together. If you have machines without internal XCVRs you just use a bnc connector to attach an XCVR to then thinnet and you are set. I much prefer thin coax over thick. USNA bought 1000' spools of rg58, strippers, connectors, and crimping tools and had no problem making all of our cable up. Our techs found that is was quicker and easier to "connectorize" rg58 then it was to to tap a thick coax segment. It is also easier to debug thinnet as you can unplug segments and terminate them any place you have an XCVR. As for having to buy a TDR, we have one which we use for both thick and thin coax. Can't run a network without one. All in all the intitial costs of wiring as above will be a bit higher than your option 3 but I think it will pay off in the long run. You will have lower maintenance costs and a lot more flexibility wiht this method. I would imagine that others will also suggest the same sort of wiring scheme. As an aside... Cabletron has also introduced a new box called a "Multi Media Access Controller" (or someting like that- I call them MMACs) which are chassis into which you mount various cards to tie your net together. There are thinnet cards, fiber cards, twisted pair cards, and maybe others (I am at home and don't have the info here- sorry). Everything is repeated to all the ports on all of the cards in the box. The MMAC also has a "brain card" which holds the master ports (2 AUI and one thinnet) as well as serial connector so you can talk to the brain. The brain comes in 2 flavors- smart and stupid- the smart brain is only a little more expensive than the stupid brain. The smart brain collects statistics and is supposed to allow the network manager to turn ports on and off and things like that. They come in 3 and 8 slot chassis. I seem to recall that a 3 slot chassis with 2 12 port thinnet repeater cards and the smart brain cost under 5k. I thought that was very cost effective since 3 MR-9000s would cost about 7200 and would also need 3 XCVRs and appropriate cables. The MMACs don't need XCVRs since they have thinnet ports on the brain card. As soon as they came out I ordered 3 of them- I never got to see them though as I left the Academy before they came in. hope this helps and good luck, Tim Smith (formerly of the US Naval Academy) 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
todds@cognos.uucp (Todd Sandor) (03/28/89)
> We're getting ready to wire our new building for ethernet. >We'd like to have five physical networks (for each group of workstations), >these networks will be linked to a backbone using Retix 2244 bridges. The >backbone will exist in the computer room only. > We have closed offices in this building, so we want one or two >ethernet taps per office. It's a new building, so adding the wiring >is easy so long as I decide in the next week or two. > I've looked into several options: > From our experience we found it best to have thick-nets (sub-nets) tied to the main backbone (via Retrix 2255's), and have multiport thin-net repeaters tied off the subnets to the individual offices. Each subnet has hosts (on sub-net backbone), with Workstations configured into these hosts (we are totally disk-less WS) tied of the multiport repeaters on that subnet. It WILL be cheaper in the long run to have one thin-net per office as if don't when have office moves you will probably require re-wiring the thinnets, and if like our company who has office moves regularly (make work projects), it will be cheaper in the long run. We use Cabletron multiports and 3-COM multiports (these allow 15 thin-net cards) as our multiport repeaters. The 3-com are the best as have the most flexibility, but delivery can take up to 2 months for this equipment (Cabletron stuff usually stock items). If have it this was, when moves happen, just go into the cable room and change the thin-nets so have correct office going off the correct multi-port (and thus sub-net), office moves are easly done, rather then having to get more equipment and cable installers each time. If I was helping in plans for a new building, I'ld be recommending each office have a thin-net tied (well labeled obviously), to a cable room, more expensive initially, but saves money and headaches later. I don't have experience with twisted pair ethernet, but you could do it in the same way. -- Todd Sandor Voice: (613) 738-1338 ext 2704 P.O. Box 9707 Cognos Incorporated FAX: (613) 738-0002 3755 Riverside Dr. uucp: todds@cognos || uunet!mitel!sce!cognos!todds Ottawa, Ontario Steady as she goes! CANADA K1G 3Z4
dfk@cwi.nl (Daniel Karrenberg) (03/29/89)
In article <170@usna.MIL> tsmith@cad.usna.mil.UUCP (Tim G. Smith) writes: > >All of the MPRs to do this were not cheap but in my opinion they were >well worth the money. If you use MPRs you can save a lot of network >debugging time- you just look for the faulted segment and then deal >with it. The whole net does not come crashing done due to one fault. >They are really great for making life easier on the network managers. >They also keep one person from screwing up the net by pulling the >network apart or taking of his terminator. If someone screws with the >parts of the net accessible in his office he only takes himslf and >maybe his officemate off the net. No one else gets affected. That is a >real nice feature when you have a lot of naive users. I agree most emphatically. We use similar setup and in the beginning it looked really expensive. But it's sure worth the ease in debugging. Of course it is essential that your MPRs indicate a faulty segment and it's very nice if they reconnect it once the fault goes away. We had early verions of the DEMPR that didn't do that so you had to go in an push the reset button. We are now using ISOLAN MPRs. They work OK and are quite dependable. We have 24 of them (one was DOA). It might also be helpful if the users could see the lights on the MPRs to find out whether their segment is broken. This might eliminate some of the calls on the network support. We don't have this however and the load is still bearable. >Wiring each office individually also makes it easier when people move >from one subnet to another (USNA subnets by departments and every >summer offices get shuffled among departments). We didn't do that because it's too expensive for us. But locating a fault in a string of say 5 offices is still quite possible and certainly better than having everything or even a complete floor stop. We also use connector boxes with 2 BNCs per connection in each office. They automatically bridge the segment at the box unless there are two BNC connectors plugged (screwed ?) in. The boxes are placed right under the ceiling out of reach of cleaning personnell and user's feets. This eliminates some faults but we had a good deal of problems with faulty microswitches in these connector boxes to start with :-). Recommendation: Use MPRs and make the segments as small as possible down to the office level if you can afford it. Keep cheapernet cables and connectors out of reach of users and other people likely to tamper. Select your MPRs carefully (indicator lights, auto reconnect, MTBF). -- Daniel Karrenberg Future Net: <dfk@cwi.nl> CWI, Amsterdam Oldie Net: mcvax!dfk The Netherlands Because It's There Net: DFK@MCVAX
rnicovic@polyslo.CalPoly.EDU (Ralph Nicovich) (03/30/89)
Tan, I read your posting to USENET about your building wiring, and I have a few sugestions. Without knowing more about the problem some of my sugestions may be inapropriate. I have just finished working on a network interconnecting some 85 buildings on a major university and have come by many of the problems you are mentioning. First off you may be selling short the capacities of a single(logical) ethernet it seems to me that you have a lot of networks tied to a backbone, you may not need so many. Another thing is that unless there are some major security concerns I would not worry to much about a user being on the propper network, just plug him in the network that is in the office, let the software find his propper resources. About the thinnet. In my opinion Thinnet is not ment to connect large groups of workstations. Although it has the same capabilities as far as bandwidth, the number of connections and the mechanical reliability of it, makes it less than ideal. With repeaters such as the CABLETRON MR9000C you can expand it but watch out for NOISE problems and to many levels of repeaters. There is no neat way you can set it up as you wish. The specs allow for 30 taps into the cable. Normaly this would be 30 "T" connectors. In my opinion if you "looped" thru an office the way you described each one would count as 3 connections. Also any person unplugging one of these cables would totaly bring down the network. I would treat the network and the wiring as two seperate resources. The networks will change as time goes by, the wiring should be done once. I would personaly wire the building with the AT&T PDS system. This entails RJ-45 jacks in the offices that "home-run" to the nearest WIRE CLOSET. I would put 2 jacks in each office, One to be used by the telephone people and the second for data. The one for the telephone could allow for up to 4 telephones. The one for data could support token-ring, 3270, rs232, ethernet etc... The only place that equipment should be is in the wire closet. Look at Ungermann-bass Access-One, at HP Starlan (10MBs version) and Cabletron. You may pay a premium on capitol equipment to start with but it will add years to your life if you have a reliable system that works. Moves and changes will be simple with patch cords. When you think about it the workstations you are connecting , if they use ethernet, are not cheap to begin with. Why make the network out of a kluge that will drasticaly effect the performance ,and usability of these same workstations? Ralph Nicovich Network Engineer Cal Poly State University, SLO (805)-756-1297 rnicovic@polyslo.calpoly.edu