ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) (04/03/91)
We're considering using our twisted pair wires for Ethernet. Hence tons of questions, some of which are below: - aside from distance/no. of stations differences, is UTP inherenly less reliable than thin/thick E-net? - in a smallish network (30 nodes, less that 300 ft.) is a passive hub OK? Price difference between active and passive hubs is substantial... - has anyone had any experience with the passive 12-port hub from Asante? - is there any advantage to "cascading" hubs (attaching the next one to a port in the previous one - you lose two ports) as opposed to putting both directly on a thin/thick backbone and using all ports for the UTP star? - (must be obvious...) does 10Base-T *have* to use star topology? If so, then (troubleshooting aspects set aside) I save maybe $500 in thinwire cabling costs but I lose $1000+ on a hub... I think that 10Base-T's economic advantages are a bit overblown given the current prices; am I right? Thanks very much for answers/opinions. E. -- Eric Behr, Illinois State University, Mathematics Department Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu
msieweke@hayes.uucp (04/03/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu>, ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: > We're considering using our twisted pair wires for Ethernet. Hence tons of > questions, some of which are below: > > - aside from distance/no. of stations differences, is UTP inherenly less > reliable than thin/thick E-net? What is reliable? Our most frequent problem with our thinwire segments is people disconnecting their equipment and not putting back the jumper. This brings down the net for everyone on the same cable. Our thickwire xceiver cables frequently come out of the back of the equipment. So far (5 months) we have had no problems with our UTP segments. If someone screws up his/her machine, no one else is affected. Also, with Synoptics equipment our UTP traffic is easier to manage. > - in a smallish network (30 nodes, less that 300 ft.) is a passive hub OK? > Price difference between active and passive hubs is substantial... I don't know the difference. I have seen TP hubs for about $150 per port. Is there another type besides this? > - (must be obvious...) does 10Base-T *have* to use star topology? > If so, then (troubleshooting aspects set aside) I save maybe $500 in > thinwire cabling costs but I lose $1000+ on a hub... I think that > 10Base-T's economic advantages are a bit overblown given the current > prices; am I right? > Eric Behr, Illinois State University, Mathematics Department > Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu Yes, 10BaseT is always physically a star. Its advantage is that you can run it off of existing wiring (if it meets spec). You save the cost of running new wire (very expensive if contracted out). You brush off the troubleshooting aspects as though they weren't a consideration. How much is your time worth? 10BaseT eliminates several potential problems and will save you time every month that you use it. It's the gift that keeps on giving. :-) In my experience only the people in a government or university are short- sighted enough to ignore the time savings. This is not intended as an insult. I worked for Georgia Tech for 7 years, and every purchase had to be justified on tangible cost. No one took the time to investigate the _people_ costs of any purchase decision. Usually quality wasn't an obvious factor in the decision either. I hope this helps. -- Mike Sieweke ...!uunet!hayes!msieweke Hayes Microcomputer Products msieweke@hayes.uucp Norcross, Georgia hayes!msieweke@uunet.uu.net
jcrowder@GroupW.cns.vt.edu (Jeff Crowder) (04/03/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: >- (must be obvious...) does 10Base-T *have* to use star topology? >If so, then (troubleshooting aspects set aside) I save maybe $500 in >thinwire cabling costs but I lose $1000+ on a hub... I think that >10Base-T's economic advantages are a bit overblown given the current >prices; am I right? I think this is a salient point. I keep wanting to go with 10Base-T for all the obvious reasons (we already have good UTP installed, the aesthetics are much better, cable fault isolation and tolerance, management, etc.). But I just can't get the numbers to add up. If you read the rags and listen to the poop, it sounds like EVERYBODY is doing 10Base-T for ALL their new installations. But its kind of funny; net managers I actually talk to aren't moving so fast. I think a lot of the hype has been invented. You know, if you're a hardware manufacturer of ethernetworking type stuff (or a seller of such), this thinwire coax thing just doesn't leave much room for revenue. Heck, you can install a small thinwire ethernet with absolutely NO electronic gizmos whatsoever. These smooth talkers and network-conference-showers have every reason to push 10Base-T as hard as they can. The good news is that this should enhance competition and lead to very rapid price incentives for customers like me. The per port cost is already on the downswing. But until it drops quite a bit more, I won't be able to justify it in most cases... Jeff Crowder Virginia Tech jcrowder@GroupW.cns.vt.edu
andrew@jhereg.osa.com (Andrew C. Esh) (04/04/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: >We're considering using our twisted pair wires for Ethernet. Hence tons of >questions, some of which are below: > >- aside from distance/no. of stations differences, is UTP inherenly less >reliable than thin/thick E-net? > UTP? Unshielded Twisted Pair, I assume. Since the inclusion of Link Pulse, I have found 10baseT to be much more reliable. One bad node doesn't affect any of the others, and the bad one is obvious. The wiring can be tested with a continuity tester, and it either works or it doesn't. Shielded wire is prone to problems that are marginal, and hard to locate without a Cable Scanner or a Time Domain Reflectometer. Connectors are much easier with 10baseT, and the distribution blocks and wiring schemes are ones that any telephone contractor can handle. >- in a smallish network (30 nodes, less that 300 ft.) is a passive hub OK? >Price difference between active and passive hubs is substantial... > >- has anyone had any experience with the passive 12-port hub from Asante? By the way, 384 feet (100 meters) is the limit for 10baseT, so 300 feet is not what I would call small. Passive and active hubs? I didn't know there was such a thing as a passive hub. Either way, the wiring is the same, so I would wire it, and then get a cheap hub on evaluation. If it runs, go for it. Be sure it is advertised as 10baseT. Ones I have tried out or helped beta test are David, Xyplex, ODS, and others. Please tell me about the active and passive deal though. This is interesting. I have no experience with Assante. > >- is there any advantage to "cascading" hubs (attaching the next one to a >port in the previous one - you lose two ports) as opposed to putting both >directly on a thin/thick backbone and using all ports for the UTP star? Lose two ports? If you are connecting two "out" ports together, you will have problems. To cascade, you need to connect an out port of the parent, to the AUI port of the child, with a 10BaseT transceiver. Without the tree heirarchy, the timing gets all crunched and one or both of the hubs will stop working until the problem is corrected. Also, I would not cascade more than two levels. Although I have not tried it, I have had experience with time delay problems with other types of hubs if there are many cascades. You need to figure a loss budget for the retransmission time, and see if it stays within the net specifications. > >- (must be obvious...) does 10Base-T *have* to use star topology? >If so, then (troubleshooting aspects set aside) I save maybe $500 in >thinwire cabling costs but I lose $1000+ on a hub... I think that >10Base-T's economic advantages are a bit overblown given the current >prices; am I right? > Yes, 10baseT has to be star, just like telephones. When you consider costs, compare the cost per foot of wire, connectors, wiring blocks, manpower for installation, and the rental or purchase of test equipment to certify that the net is correctly installed. A Pair Tester is a lot cheaper than a Cable Scanner, and a four year old can operate it. Maintenance should also be figured. I have found Thin net to be prone to connector trouble after a few months getting kicked around under the desk. UTP is less sensitive to this sort of thing. >Thanks very much for answers/opinions. E. >-- >Eric Behr, Illinois State University, Mathematics Department >Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Minneapolis, MN 55416-1528 So much System, (612) 525-0000 so little CPU time...
jbreeden@netcom.COM (John Breeden) (04/04/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: >We're considering using our twisted pair wires for Ethernet. Hence tons of >questions, some of which are below: > >- aside from distance/no. of stations differences, is UTP inherenly less >reliable than thin/thick E-net? > No (I can hook up just as many stations using UDP as thin/thicknet). >- in a smallish network (30 nodes, less that 300 ft.) is a passive hub OK? >Price difference between active and passive hubs is substantial... > If you are using a "passive" hub, it's not 10baseT. The standard calls for hubs that are active repeaters. If you decide to use "passive" hubs, you'll save money but a) you're not 10baseT and b) you're not "reliable". One of the things that makes 10baseT reliable is the active nature of the hubs (ie: multiport repeater function - packet re-generation). > >- is there any advantage to "cascading" hubs (attaching the next one to a >port in the previous one - you lose two ports) as opposed to putting both >directly on a thin/thick backbone and using all ports for the UTP star? > No, you can attach hubs off of coax, chain hub to hub or star hubs off of a single hub or any combination. >- (must be obvious...) does 10Base-T *have* to use star topology? >If so, then (troubleshooting aspects set aside) I save maybe $500 in >thinwire cabling costs but I lose $1000+ on a hub... I think that >10Base-T's economic advantages are a bit overblown given the current >prices; am I right? > Yes, 10baseT must be in a star configuration (it's in the standard). There is nothing that stops you from using a bus configuration (other than mechanical, you need to flip wire pairs to make it work). You loose the "reliability" of 10baseT without the hubs (packet regeneration, port jamming on bad packets, station isolation, "added-value vendor" hub management etc). 10baseT's economic advantage is'nt in the installation (that rumor got started with the "cost savings" of using existing phone wiring, which considering the type and condition of the wire installed may not be the case. The "cost savings" is the ease of trouble reporting, isolation and management due to the a) star configuration and b) management and fault isolation inherent in the hubs. If you are looking for "cheap" - use thinnet. -- John Robert Breeden, jbreeden@netcom.com, apple!netcom!jbreeden, ATTMAIL:!jbreeden ------------------------------------------------------------------- "The nice thing about standards is that you have so many to choose from. If you don't like any of them, you just wait for next year's model."
ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) (04/04/91)
First, thanks to all who responded - I'm trying to acknowledge all replies, but... andrew@jhereg.osa.com (Andrew C. Esh) writes: >In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: > >Passive and active hubs? I didn't know there was such a thing as a passive >hub. This is a result of my ignorance; let me explain. Asante's ad caught my eye the other day; I called them, found that the hub is cheap, so I asked a tech type if it does all that others do - reclocking signals etc. He said "no, it's a passive hub" and of course I took his expertise for granted. So there. Maybe he didn't know what he was talking about? >Lose two ports? If you are connecting two "out" ports together, you will >have problems. Most hubs I looked at had one out port configurable as a cascade port. >-- >Andrew C. Esh andrew@osa.com >Open Systems Architects, Inc. Thanks! E. -- Eric Behr, Illinois State University, Mathematics Department Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu
asteiner@casbah.acns.nwu.edu (Albert Steiner) (04/04/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: > - has anyone had any experience with the passive 12-port hub from Asante? No experience with Asante, but Asante, SMC, David Volksnet and some other cheap (<$1000) hubs have no management as well as other cutting other corners. It's not that they are passive. ----------------------------------------- asteiner@casbah.acns.nwu.edu Albert Steiner, Academic Computing and Networking, Northwestern University, Evanston, IL 60208 708-491-4056
dan@gacvx2.gac.edu (04/04/91)
In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu>, ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: > We're considering using our twisted pair wires for Ethernet. Hence tons of > questions, some of which are below: > > - aside from distance/no. of stations differences, is UTP inherenly less > reliable than thin/thick E-net? I have found UTP to be more reliable than thin ethernet. We have been using thin ethernet in several building. I have found that the BNC connectors on thin ethernet do not take wear well. I have had faculty rearrange their office and take out an entire segment consisting of several offices, not because they are being malicious but just because they don't know better. In classrooms I use a $75 connector and cable from AMP that shorts out inside the wall when the cable is removed. Most faculty offices have a $8 wall plate with two BNC connectors on it. The real solutions are to use more expensive connectors, or change to a star topology (DEC recommends this in their local area networking books.) I have also found that the physics department likes to extend their cable on their own. When I finally caught up with them they had a network that was quite a bit over the 30 connections and 300 meters allowed in a thin ethernet. A unix computer on their segment could not be reliably accessed by users on other segments. The physics departments excuse was that I didn't make the rules clear enough and just how were they suppose to know the length of the cable. They still paid for the repeater and earned my ire. The faculty in the building I have using UTP ethernet have not been able to mess up the network yet. I have had some stations go bad in the lab due to strained cables, but a longer cable took care of that. > - in a smallish network (30 nodes, less that 300 ft.) is a passive hub OK? > Price difference between active and passive hubs is substantial... > > - has anyone had any experience with the passive 12-port hub from Asante? I don't know that active and passive hubs is a good way to describe the difference between the types of hubs. All hubs are essentially multi-port repeaters some have incorporated bridges or a high speed internal bus. The more expensive hubs usually include managment capabilities. This allows monitoring all the ports from one workstation anywhere on the ethernet. Most managment hubs allow the managment station to get the number of packets/bytes transfered through the port and also an error count. Many allow the management station to turn ports on and off. I have built a 60 computer network using the Asante hubs. They work well. They have no management capability. I did run into some problems with the Asante hub not being able to tell a shorted cable from a cable with a computer at the other end. I was told by another use of the hubs that this is because Asante used a single chip UTP ethernet controler that was designed for workstation and as uses the same link integrity pulse as a workstation, thus the loopback caused by the short could not be detected. I don't know if other hubs have the same problem, if they do then something more than a single LED is needed to signal an error. Having managment would help. I also had a problem with a 25 pair cable that was multipled to another punch block this cable caused the hub to go throught a cycle of powerup-error-reset-error-reset-... The whole hub did this making all the workstations on it unuseable. I have seen token ring hubs blow fuses in similar situation. More expensive hubs just cut out the malfunctioning ports. Fixing the cabling problem caused the hub to function normally. In a small network the Asante hub is adiquate. I do not plan to use it in the new Computer Science/Physics/Academic Computing building. There are more than 200 computers in this new building on three floors to start with, the building it being wired by the same contractor that is doing the phones and is going to be capable of supporting about 500 computers before we need to add wire. I am planning to use management hubs in this building. There is also significant need for traffic management in this building. > - is there any advantage to "cascading" hubs (attaching the next one to a > port in the previous one - you lose two ports) as opposed to putting both > directly on a thin/thick backbone and using all ports for the UTP star? I use thin ethernet between the hubs. This reduces the repeater count and allows me to use the cascading feature in a room with one UTP port. I can plug in a hub and have 11. With RJ45 cords a quick network can be build in any classroom. We will be using this this summer as the summer programs here will out number our available computer labs since three labs will be closed while they are moving to the new building. I is also hand if the instructor wants to bring more than one computer into a classroom. Even in the new building I will probably have a couple of Asante hubs available for this use. > - (must be obvious...) does 10Base-T *have* to use star topology? > If so, then (troubleshooting aspects set aside) I save maybe $500 in > thinwire cabling costs but I lose $1000+ on a hub... I think that > 10Base-T's economic advantages are a bit overblown given the current > prices; am I right? I compare 10BASE-T to a thin ethernet star. A thin eithernet bus does not have the same reliability characteristics as a star. It does have to be a strict star where the thin ethernet star can have multiple computers on each segment. I think you will find multi-port thin ethernet repeaters to as expensive as UTP hubs. I also use the UTP wiring for more than just ethernet, we also run token ring and rs-232 over it. It is also possible to run AS/400 twinax terminals and 3270 terminals over it using baluns. Arcnet is another possibility. I have found that a well built thin ethernet network and a UTP ethernet are about the same price per port. By well built I mean one with about the same reliability as the UTP network. I also strongly recommend purchasing a portable cable tester like the ones sold by MicroTest. They have been valuable in finding both thin ethernet and UTP cable problems. Also a cheap PC software package like Ether Monitor from Brightworks or NetCure/NetMon (available on the net) is very helpfull for locating problems. -- Dan Boehlke Internet: dan@gac.edu Campus Network Manager BITNET: dan@gacvax1.bitnet Gustavus Adolphus College St. Peter, MN 56082 USA Phone: (507)933-7596
andrew@jhereg.osa.com (Andrew C. Esh) (04/04/91)
In article <3901.27f9e38a@hayes.uucp> msieweke@hayes.uucp writes: > >What is reliable? Our most frequent problem with our thinwire segments is >people disconnecting their equipment and not putting back the jumper. This > Not putting back the jumper? I have something which may help. Tell the users to disconnect the whole T-connector, and leave it on the wire. I have done this many times, and the net is just fine. They don't need to replace the T with a straight thru barrel connector, which is what I think you mean by "jumper". This practice may cut down on the amount of inexperienced alteration of the physical characteristics of your net. :-) >-- >Mike Sieweke ...!uunet!hayes!msieweke >Hayes Microcomputer Products msieweke@hayes.uucp >Norcross, Georgia hayes!msieweke@uunet.uu.net -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Minneapolis, MN 55416-1528 (612) 525-0000 Practicing the OSI Standard
phil@shl.com (Phil Trubey) (04/09/91)
In article <1548@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >In article <1991Apr03.004515.12021@rs6000.cmp.ilstu.edu> ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) writes: >>- (must be obvious...) does 10Base-T *have* to use star topology? >>If so, then (troubleshooting aspects set aside) I save maybe $500 in >>thinwire cabling costs but I lose $1000+ on a hub... I think that >>10Base-T's economic advantages are a bit overblown given the current >>prices; am I right? > >I think this is a salient point. I keep wanting to go with 10Base-T for >all the obvious reasons (we already have good UTP installed, the >aesthetics are much better, cable fault isolation and tolerance, >management, etc.). But I just can't get the numbers to add up. > >If you read the rags and listen to the poop, it sounds like EVERYBODY is >doing 10Base-T for ALL their new installations. But its kind of funny; >net managers I actually talk to aren't moving so fast. I think a lot of the >hype has been invented. You know, if you're a hardware manufacturer of >ethernetworking type stuff (or a seller of such), this thinwire coax >thing just doesn't leave much room for revenue. Heck, you can install a >small thinwire ethernet with absolutely NO electronic gizmos whatsoever. >These smooth talkers and network-conference-showers have every reason to >push 10Base-T as hard as they can. The good news is that this should >enhance competition and lead to very rapid price incentives for >customers like me. As others have pointed out, the cost savings aren't in physical capital costs - they are in people time saved from not having to fix and diagnose broken segments. This is a *major* cost for any network over a few dozen nodes. Add to this the fact that the dozens of LAN users won't experience network outages and you have large cost *savings* going to 10BaseT for most environments. As fas as the hype being invented, every new installation that I've dealt with (I work for a systems integrator) in the last year has been 10BaseT. The *only* place where I would recommend coax is in a lab environment where the machines are close together and don't move (ever), in a computer room connecting a few hosts together, or for a demo setup where you want to connect a few machines together for an on site demo. -- Phil Trubey | Internet: phil@shl.com SHL Systemhouse Inc. | UUCP: ...!uunet!shl!phil 50 O'Connor St., Suite 501 | Phone: 613-236-6604 x667 Ottawa, Ontario, Canada | Fax: 613-236-2043
jcrowder@GroupW.cns.vt.edu (Jeff Crowder) (04/09/91)
In article <1991Apr08.171237.19978@shl.com> phil@shl.com (Phil Trubey) writes: >>>thinwire cabling costs but I lose $1000+ on a hub... I think that >>>10Base-T's economic advantages are a bit overblown given the current >>>prices; am I right? >> >>I think this is a salient point. I keep wanting to go with 10Base-T for >>all the obvious reasons (we already have good UTP installed, the >>aesthetics are much better, cable fault isolation and tolerance, >>management, etc.). But I just can't get the numbers to add up. >As others have pointed out, the cost savings aren't in physical >capital costs - they are in people time saved from not having to >fix and diagnose broken segments. This is a *major* cost for any >network over a few dozen nodes. Add to this the fact that the dozens Well, I can appreciate that a manageable hub should provide better fault isolation and diagnostic capabilities. In real terms, however, we've had only 3 physical layer failures which required dispatching a diagnostician within the last 14 months on a campus wide network connecting several thousand machines. Thinwire coax *installed properly* (i.e. good stress relief and careful connector installation) is quite reliable. Of course, a modicum of user training is recommended; it does not help to have a geology professor unplug the cable from the tee on his machine. I might also mention that the vast majority of responses I've received via e-mail have indicated that the authors had already come the more or less the same conclusion I had. (threads of paranoia showing thru) >As fas as the hype being invented, every new installation that I've dealt >with (I work for a systems integrator) in the last year has been 10BaseT. >The *only* place where I would recommend coax is in a lab environment Do you SELL the hubs you install by any chance ... ??? In any event, I'll bet you haven't helped out with many installations at state supported universities in states where very scarce funds are being channelled out of education and into a presidential campaign fund... >Phil Trubey | Internet: phil@shl.com Jeff Crowder, Network Guy and Grass Mower Virginia Tech jcrowder@GroupW.cns.vt.edu
phil@shl.com (Phil Trubey) (04/09/91)
In article <1582@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >>As fas as the hype being invented, every new installation that I've dealt >>with (I work for a systems integrator) in the last year has been 10BaseT. >>The *only* place where I would recommend coax is in a lab environment > >Do you SELL the hubs you install by any chance ... ??? Yes, however we don't have exclusive distributorships with any particular product - we can markup Ethernet transceivers as well as we can 10BaseT hubs (on fixed price contracts). Interestingly enough, the reason for the majority of our 10BaseT installations is that our customers *ask* for it. > In any event, >I'll bet you haven't helped out with many installations at state >supported universities in states where very scarce funds are being >channelled out of education and into a presidential campaign fund... Stripped of the jabs, you're quite right. Universities tend to view labour as cheap and capital equipment as expensive. Most operating companies operate exactly the opposite way 'round. At least that's what I've found after working in both environments... -- Phil Trubey | Internet: phil@shl.com SHL Systemhouse Inc. | UUCP: ...!uunet!shl!phil 50 O'Connor St., Suite 501 | Phone: 613-236-6604 x667 Ottawa, Ontario, Canada | Fax: 613-236-2043
morgan@Panther.Stanford.EDU (RL "Bob" Morgan) (04/10/91)
Re 10Base-T: >As others have pointed out, the cost savings aren't in physical >capital costs - they are in people time saved from not having to >fix and diagnose broken segments. Hmm, my numbers show that 10Base-T is cheaper in almost every case (excepting those special cases of student labs and such that benefit from big-time daisy-chaining), even before the less immediate wins of using the structured cabling, etc. The cheapest thinnet MPR we've found has 9 ports (8 thin + 1 AUI) and runs about $1300 = $162.50 per thinnet port. The cheapest 10Base-T MPR has 14 ports (12 T + 1 thin + 1 AUI) and they want to sell it to us for the remarkable price of $700 = $58.33 per T port. A $100 win per port right off the bat. Lest you think this is a fluke, there are multiple vendors beating down our doors wanting to sell 10Base-T MPRs for around $900/12 = $75/port. Now cabling. My rule of thumb for contractor-installed thinnet in existing office space is about $100/office for quick & dirty, $150 for nice-looking (ie all in wiremold, etc). This isn't really affected much by whether you daisy-chain or not, since most of the cost is in the connectors and the time spent in the rooms and stringing the cable, not in the cable itself. Our on-campus telephone people, who support their operation entirely by user fees, are willing to do the wiring part of installing 10Base-T (ie, mounting punchblocks and cross-connects and labelling and any futzing with the jack that might be required) for $50 per circuit plus a $150 setup per job. We also choose to use a 110 block with RJ-45s mounted in it for jumpering to the MPR, which runs $100 or so. BTW, I haven't found an existing telephone circuit here yet that couldn't run 10Base-T acceptably (but then we haven't put in that many yet). 10Base-T Ethernet cards are only a few dollars more than thinnet these days. Jumper cables (wall outlet to computer) cost about the same for the two methods. So, comparing variable costs for an 8-station network (which puts thinnet in the best light): Thin: Repeater: $1300 Cable installation: 800 ---- 2100 10Base-T: Repeater: 700 Cable installation: 650 ---- 1350 So 10Base-T wins already, not even considering that we have 4 ports left over, we didn't have to get contractor bids, users didn't have to get their offices disrupted, moves and changes are much easier, etc. Even if all this weren't already compelling enough, an enormous win in our situation is that our on-campus telephone people are prepared to install these kinds of nets using their existing techs who already know how to handle twisted pair cable. This simply wasn't the case for thinnet. We're working out the procedures now so they can just take the whole process and run with it. So the *biggest* cost savings of all (IMHO) is that these networks can be installed en masse by readily-available, well-trained telephone people; you no longer need to use hard-to-find, expensive computer-network types who'd rather spend all their time sending news articles 8^). I'd like to see the numbers of those who say thinnet is cheaper. Again, I'll believe you if you're mostly installing student labs and such, but here at least those are mostly done already. The challenge is giving network access to every single office on campus, and that's where 10Base-T wins big. - RL "Bob" Morgan Networking Systems Stanford
andrew@jhereg.osa.com (Andrew C. Esh) (04/10/91)
In article <1582@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >In article <1991Apr08.171237.19978@shl.com> phil@shl.com (Phil Trubey) writes: > >>As fas as the hype being invented, every new installation that I've dealt >>with (I work for a systems integrator) in the last year has been 10BaseT. >>The *only* place where I would recommend coax is in a lab environment Only a lab environment? You must realize that some sites are more spread out than what 10baseT can reach. Sure, it will work for one floor of a small to medium sized building, but sheer physics drives you up into ThinNet, ThickNet, and Fiber when the physical seperation between nodes is great. 10baseT just doesn't go more than 400 ft. The company I work for is called upon to design nets for buildings up to a quarter MILE on a side. 10baseT is great for an office area, but you need something else for distance. > >Do you SELL the hubs you install by any chance ... ??? In any event, >I'll bet you haven't helped out with many installations at state >supported universities in states where very scarce funds are being >channelled out of education and into a presidential campaign fund... > >>Phil Trubey | Internet: phil@shl.com > >Jeff Crowder, Network Guy and Grass Mower >Virginia Tech >jcrowder@GroupW.cns.vt.edu The company I work for is "vendor inspecific" for just this reason. :-) -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Minneapolis, MN 55416-1528 (612) 525-0000 Practicing the OSI Standard
nelson@sun.soe.clarkson.edu (Russ Nelson) (04/10/91)
In article <1991Apr9.181721.15560@leland.Stanford.EDU> morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes:
Thin:
Repeater: $1300
Cable installation: 800
----
2100
10Base-T:
Repeater: 700
Cable installation: 650
----
1350
Perhaps I'm being too stupid here, but why do you need a repeater for
an 8-station network? Why not hook them all up to the same cable? In
that case, you have $800 for Thin vs $1350 for 10Base-T.
--
--russ <nelson@clutx.clarkson.edu> I'm proud to be a humble Quaker.
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.morgan@Panther.Stanford.EDU (RL "Bob" Morgan) (04/10/91)
> Perhaps I'm being too stupid here, but why do you need a repeater > for an 8-station network? Why not hook them all up to the same > cable? Well, if you've been reading comp.dcom.lans, the benefits of structured wiring have been explained at length. But this is indeed just the sort of question that those of us who install these things for a living have to answer on a daily basis. You're right, it isn't obvious. So here are the answers in easy-to-understand terms. 1. I didn't say so in my previous note, but the main point of putting in a network around here (and at almost any large institution these days, I'll bet) isn't to hook the machines to themselves, it's to hook them to the campus net (that is, to the Internet). So your proposed single run of thinnet has to connect to the building backbone in any case, which means a repeater. I'll admit I can't explain it, but I can get a 12-port 10Base-T repeater these days for about the same price as a traditional 2-port AUI-only repeater. 2. Where are these 8 stations anyway? As I said before, if they're in one room, sure, use thinnet. But if, as is the usual case, they're scattered around the floor of your medium-to-large building, then the "single cable" ends up hopping from room to room, crossing the hall, going around the corner, and doubling back to catch the one at the end of the hall. Maybe you're lucky and it's still only 450 feet. Then they get two more computers in two different offices (and yes, they *always* get more computers eventually). Someone looks up in the ceiling and finds where the cable runs kind of near the one office, cuts it and adds a loop. The other one is sort of near the end of the daisy-chain, so they just add to the end. How long is the cable now? Where are its connectors that will fail over time? Who bought the new cable that was spliced in, and why didn't they notice that it was 53 ohms, not 50 (they're lucky it wasn't 75)? So eventually the thing breaks down, and after weeks of cursing and head-scratching, they call in the network guys, who look around, shake their heads, and say: what you should do is buy a MPR and install some home runs to it. That way you'll have a manageable cable plant that is relatively immune to the weird things that people do to nets, and that provides an obvious growth path for future installers. 3. If those 8 stations are typical office PCs these days, they're probably worth a total of $20,000 or so. If they're fancy Unix boxes, the total may be over $100,000. If the people using them are getting paid for their time, the total may be a couple of hundred dollars an hour (of course, if they're just grad students ... 8^). Isn't it worth spending a extra few hundred dollars, adding at most 3 or 4 percent to the total cost of this networked computer system, to make the network as reliable as it can possibly be? So, the point is: no matter whether you're installing thinnet or twisted-pair or fishing line, it's almost always worth it to install a structured wiring plant. And once you've decided to do that, *if* you can use your existing structured telephone cable plant, then 10Base-T is cheaper by a long shot. No hype. - RL "Bob" Morgan Networking Systems Stanford
brinich@keinstr.uucp (Mark Brinich) (04/10/91)
In article <1582@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: > >Well, I can appreciate that a manageable hub should provide better fault >isolation and diagnostic capabilities. In real terms, however, >we've had only 3 physical layer failures which required dispatching a >diagnostician within the last 14 months on a campus wide network >connecting several thousand machines. Thinwire coax *installed >properly* (i.e. good stress relief and careful connector installation) >is quite reliable. Of course, a modicum of user training is >recommended; it does not help to have a geology professor unplug the >cable from the tee on his machine. > I thought I'd throw in my 2 cents worth, as I have looked at this issue both at my current position, and in my previous life. Given budget constraints, it seems that coax, properly installed, is the way to go. In addition to this AMP (along with Black Box, DEC, etc.)has a very good wiring system that does away with all the bad traits of coax. With this wiring scheme you can unplug your Ethernet cable either at the wall connector they supply or at your workstation without bringing down your ethernet segm ent. Properly installed this also eliminates user access to the coax itself. Thus you have eliminated all the bad traits of thin coax. UTP still has the advantage of isolating a problem quickly, but with AMP's wiring system you still have the capability of removing a station without affecting the rest of the Ethernet, just as you do with UTP. I'd be a full supporter of UTP if it weren't so expensive/node for equipment. In fact if I were putting in new cabling in a building, I'd put in both thin and UT P. That way if UTP ever comes down in price, you can use it. But of course by the time that happens 10mbit/sec is going to be too slow for everyone, and you'll be looking at something else by then. -- Mark Brinich voice mail(or maybe the real live thing)216 498-2821 e-mail uunet!keinstr!brinich Keithley Instruments 28775 Aurora Rd. Cleveland, Ohio 44139-1891
german@uxh.cso.uiuc.edu (Gregory German) (04/10/91)
morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes: >> Perhaps I'm being too stupid here, but why do you need a repeater >> for an 8-station network? Why not hook them all up to the same >> cable? >Well, if you've been reading comp.dcom.lans, the benefits of >structured wiring have been explained at length. But this is indeed >just the sort of question that those of us who install these things >for a living have to answer on a daily basis. You're right, it isn't >obvious. So here are the answers in easy-to-understand terms. >1. I didn't say so in my previous note, but the main point of putting >in a network around here (and at almost any large institution these >days, I'll bet) isn't to hook the machines to themselves, it's to hook >them to the campus net (that is, to the Internet). So your proposed >single run of thinnet has to connect to the building backbone in any >case, which means a repeater. I'll admit I can't explain it, but I >can get a 12-port 10Base-T repeater these days for about the same >price as a traditional 2-port AUI-only repeater. >2. Where are these 8 stations anyway? As I said before, if they're >in one room, sure, use thinnet. But if, as is the usual case, they're >scattered around the floor of your medium-to-large building, then the >"single cable" ends up hopping from room to room, crossing the hall, >going around the corner, and doubling back to catch the one at the end >of the hall. Maybe you're lucky and it's still only 450 feet. >Then they get two more computers in two different offices (and yes, >they *always* get more computers eventually). Someone looks up in the >ceiling and finds where the cable runs kind of near the one office, >cuts it and adds a loop. The other one is sort of near the end of the >daisy-chain, so they just add to the end. How long is the cable now? >Where are its connectors that will fail over time? Who bought the new >cable that was spliced in, and why didn't they notice that it was 53 >ohms, not 50 (they're lucky it wasn't 75)? If the UTP wire already pulled is useable then it is a simple matter to "pre-wire" every phone location. If you use a slotted concentrator you only have to add enough ports to meet your current needs and adding ports as needed. This is a big win over either pulling coax to EVERY location where you might want a computer or having a contractor come in and add cable on a case by case basis which will end up costing you more per connection than doing it all at once. >So eventually the thing breaks down, and after weeks of cursing and >head-scratching, they call in the network guys, who look around, shake >their heads, and say: what you should do is buy a MPR and install some >home runs to it. That way you'll have a manageable cable plant that >is relatively immune to the weird things that people do to nets, and >that provides an obvious growth path for future installers. I concur with Bob on this. The more networks you install, the more you will realize that the cost of doing it right outweighs the cost of the time lost. While it is true that the end users will rarely (if ever) give you a second thought if your well designed network does its job there is no quesiton who they will blame if the network is down because of cable problems. >3. If those 8 stations are typical office PCs these days, they're >probably worth a total of $20,000 or so. If they're fancy Unix boxes, >the total may be over $100,000. If the people using them are getting >paid for their time, the total may be a couple of hundred dollars an >hour (of course, if they're just grad students ... 8^). Isn't it >worth spending a extra few hundred dollars, adding at most 3 or 4 >percent to the total cost of this networked computer system, to make >the network as reliable as it can possibly be? >So, the point is: no matter whether you're installing thinnet or >twisted-pair or fishing line, it's almost always worth it to install >a structured wiring plant. And once you've decided to do that, *if* >you can use your existing structured telephone cable plant, then >10Base-T is cheaper by a long shot. No hype. In those cases where I do not have a useable UTP cable plant for 10baseT I still use thinnet, but in most every case every office/lab will have its own port on a multi-port repeater. Flexibility and reliability are my prime motivating factors. Each office is on their own home run and somewhat isolated from the "usual" problems of the daisy-chained environment and switching them from one network to another is as simple as choosing which repeater they are connected to in the closet. UTP is even more flexible since I can pre-wire 2 pair from each office to a patch pannel and then use that wire for 10baseT, TRN, AppleTalk, 3270 or serial connetions. I am finding out that I have users on Macs that start out on AppleTalk and then upgrade to ethernet. All I have to do in most of my networks is remove the jumper cable between the StarController and the patch pannel associated with that office and replace it with one to a 10baseT hub. Even though we don't use it in most cases I have also designed in the ability to add intelligent monitoring tools to the the hubs. LANview with Cabletron repeaters and an module with SNMP capability in the Plexcom chasis. I expect to see increased use of these as tools become available. > - RL "Bob" Morgan > Networking Systems > Stanford -- Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) US Mail: Univ of Illinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801 Office: 129 Digital Computer Lab., Network Design Office
morgan@Panther.Stanford.EDU (RL "Bob" Morgan) (04/11/91)
> Given budget constraints, it seems that coax, properly installed, > is the way to go. In addition to this AMP (along with Black Box, > DEC, etc.)has a very good wiring system that does away with all the > bad traits of coax. Hmm, I don't remember the price of the AMP system exactly, but as I recall it was on the order of $50/outlet for the fancy connectors and patch cables. Given budget constraints, who would pay to have this put in? Given that 10Base-T repeaters are $60/port, where's the cost savings? - RL "Bob"
henry@zoo.toronto.edu (Henry Spencer) (04/11/91)
In article <1991Apr10.172643.26334@leland.Stanford.EDU> morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes: >Hmm, I don't remember the price of the AMP system exactly, but as I >recall it was on the order of $50/outlet for the fancy connectors and >patch cables... We looked at it, briefly, and decided it was unattractive. Too expensive, and of course the length of the patch cable counts double against your thinwire length limit (a non-trivial concern here because we've got a big building with networking enthusiasts thinly scattered). It's a cheap (in both senses of the word) imitation of having a thickwire cable with transceivers clamped on and transceiver cables running out to customers. I increasingly think that thinwire makes sense for wiring a densely-populated room or a minimal configuration, but not for large-scale building networks. You want either thick coax with transceivers or UTP. -- And the bean-counter replied, | Henry Spencer @ U of Toronto Zoology "beans are more important". | henry@zoo.toronto.edu utzoo!henry
jcrowder@GroupW.cns.vt.edu (Jeff Crowder) (04/11/91)
In article <1991Apr9.181721.15560@leland.Stanford.EDU> morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes: > >Re 10Base-T: >Hmm, my numbers show that 10Base-T is cheaper in almost every case >The cheapest thinnet MPR we've found has 9 ports (8 thin + 1 AUI) and > ... > (a bunch of strange mumblings) > ... >So, comparing variable costs for an 8-station network (which puts >thinnet in the best light): > >Thin: >Repeater: $1300 >Cable installation: 800 > ---- > 2100 >10Base-T: >Repeater: 700 >Cable installation: 650 > ---- > 1350 > >So 10Base-T wins already, not even considering that we have 4 ports >Even if all this weren't already compelling enough, an enormous win in >our situation is that our on-campus telephone people are prepared to >install these kinds of nets using their existing techs who already >know how to handle twisted pair cable. This simply wasn't the case >for thinnet. We're working out the procedures now so they can just (Sorry if this thread is getting wearisome but I think this warrants response...) Uh, excuse me Bob but what the hell are you talking about? Of COURSE you daisy chain offices together with thinwire. I mean, after all you have more than 600' to play with. You'd have to be NUTS to dedicate each port on a thinwire MR to a single user! (Unless of course you have bucks to throw away which I doubt many of us do.) We probably average 10 hosts on a segment. Plug THAT into your numbers. And your telephone installers must be total losers if they can't handle coaxial cable at all. My installers picked it up in about 5 minutes. The designers took a bit longer, maybe 15 minutes. We've had the process integrated into the work order infrastructure for quite a while. And its nit picking but I take exception to your assertion that utp is actually cheaper to install. I can buy good thinwire for about 8 cents per foot and level 4 UTP for about 10 cents per foot. I know you don't need level 4 cable for ethernet but I wouldn't install anything else these days. I figure it costs me about $130 per workstation to do a nice job with either. EXCEPT where I DO have hosts colocated, I can do the marginal unit for next to nothing with thinwire. With UTP I have to go a whole extra cable and port OR buy one of those converter units. If you want, I'll forward a some spreadsheets done for actual network comparisons. Wow... Jeff Crowder Virginia Tech jcrowder@GroupW.cns.vt.edu
german@uxh.cso.uiuc.edu (Gregory German) (04/11/91)
jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >In article <1991Apr9.181721.15560@leland.Stanford.EDU> morgan@Panther.Stanford.EDU (RL "Bob" Morgan) writes: >> Bob's case for UTP deleted. >Uh, excuse me Bob but what the hell are you talking about? >Of COURSE you daisy chain offices together with thinwire. I mean, after >all you have more than 600' to play with. You'd have to be NUTS to >dedicate each port on a thinwire MR to a single user! (Unless of course >you have bucks to throw away which I doubt many of us do.) >We probably average 10 hosts on a segment. Plug THAT into your >numbers. Money is not the only issue here. For the most part I would claim that it is asking for trouble to daisy-chain between offices and I very rarely design a network that includes daisy-chains outside of labs/classrooms. Your users are at the mercy of the guy in the next office and everyone on that segement MUST be on the same network. If you want to place one person on a separate network you have to rewire. >And your telephone installers must be total losers if they can't handle >coaxial cable at all. My installers picked it up in about 5 minutes. >The designers took a bit longer, maybe 15 minutes. We've had the >process integrated into the work order infrastructure for quite a >while. It is hard to find people are always consistant at making good cables no matter what the type. This is one of the most frustrating problems to have to deal with over and over again and no cable type is immune to it. >And its nit picking but I take exception to your assertion that utp is >actually cheaper to install. I can buy good thinwire for about 8 cents >per foot and level 4 UTP for about 10 cents per foot. I know you don't >need level 4 cable for ethernet but I wouldn't install anything else >these days. I figure it costs me about $130 per workstation to do a >nice job with either. EXCEPT where I DO have hosts colocated, I can do >the marginal unit for next to nothing with thinwire. With UTP I have to >go a whole extra cable and port OR buy one of those converter units. I don't think anyone is really advocating pulling UTP. My main reason to use it is to avoid having to pull cable into each office right next to unused UTP pairs. If I have to install cable I would use thinnet, BUT I would for the most part opt for homeruns for reliability, flexibility and manageability. You have a valid point about colocated hosts being a problem with 10baseT. The converters back to coax can work, but are in the $450 range and have some limitations. My point is that there are other considerations than just cost and other costs than just installation. IMHO there is and will continue to be a place for both technologies. >If you want, I'll forward a some spreadsheets done for actual >network comparisons. >Wow... >Jeff Crowder >Virginia Tech >jcrowder@GroupW.cns.vt.edu -- Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) US Mail: Univ of Illinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801 Office: 129 Digital Computer Lab., Network Design Office
pat@hprnd.rose.hp.com (Pat Thaler) (04/12/91)
> > > >- is there any advantage to "cascading" hubs (attaching the next one to a > >port in the previous one - you lose two ports) as opposed to putting both > >directly on a thin/thick backbone and using all ports for the UTP star? > Lose two ports? If you are connecting two "out" ports together, you will > have problems. To cascade, you need to connect an out port of the parent, > to the AUI port of the child, with a 10BaseT transceiver. Without the tree > heirarchy, the timing gets all crunched and one or both of the hubs will > stop working until the problem is corrected. There seems to be some confusion here. Perhaps it's because 1BASE5 (1 mb/s twisted pair version of IEEE 802.3, usually called StarLAN) had hubs it a special port to cascade to the next higher hub. In any 10 mb/s IEEE 802.3 repeater, any port can be connected to another repeater port. Repeater operation, timing, etc are the same for all ports. There is one thing you should check when connection repeaters together by their twisted pair ports. Pins 1 & 2 are the TX pair for DTEs and pins 3 & 5 are the Rx pair for DTEs. Inorder to allow wiring straight through without crossing the wire pairs, repeaters normally include an internal crossover function. That is, they receive on pins 1 & 2 and transmit on pins 3 & 5. When connecting repeaters together, you need to get one repeater's transmitter connected to the other's receiver. You do this either by swapping the wire pairs, by changing a switch in the repeater which swaps the pairs internally (removes the crossover function), or by using an external MAU connected to an AUI port. The IEEE 802.3 standard allows up to 4 repeaters between any two nodes in a collision domain (ie not seperated by bridges). Some vendors will support more repeaters in certain configurations and there has been some discussion in the 802.3 Working Group of including more flexible configuration rules within a future addition to IEEE 802.3. Pat Thaler
woody@ucscb.UCSC.EDU (Bill Woodcock) (04/12/91)
> Hmm, I don't remember the price of the AMP > system exactly, but as I recall it was on the > order of $50/outlet for the fancy connectors > and patch cables. Given budget constraints, > who would pay to have this put in? Given that > 10Base-T repeaters are $60/port, where's the > cost savings? I put in a mid-sized net using the AMP hardware, and cost of the faceplate, drop cable, BNC "FastTap" jack, two RJ-13s, an RJ-45, and the gangbox came to nearly $70 per location. Nice stuff, though, and easy to install. I'm doing a roughly comparable job at the moment, using Leviton's InfoQuad jacks. They're putting them together for me custom, and the whole thing comes to just under $20 each. Not bad for a custom order on only 36 units. -Bill Woodcock BMUG NetAdmin _______________________________________________________________________________ 0000 : 0600 0800 7700 FE00 FF 0 FF8 7F 0 3 00 48 bill.woodcock.iv 0010 : CC00 7C00 0C00 1000 2800 440 8200 40 0020 : C000 4000 8000 C000 C 00 800 80 48 0 9 woody@ucscb.ucsc.edu 0030 : FF00 F000 7F80 CC00 CC 0 7 80 7 80 CC 6 0040 : CC00 7F80 9800 7800 CC00 CC0 CC 0 C 00 2355.virginia.st 0050 : CC00 CC00 FC00 CC00 CC 0 2000 10 0 7 00 C 0 0060 : CC00 CC00 CCC0 4800 2 00 1 00 2 00 48 0 9 berkeley.california 0070 : 4800 2400 1200 1000 2800 4 00 8 00 E0 0080 : 3000 6000 F000 6000 000 60 0 600 C0 0 01C 94709.1315 _______________________________________________________________________________
cmilono@netcom.COM (Carlo Milono) (04/12/91)
In article <1991Apr9.221136.12326@jhereg.osa.com> andrew@jhereg.osa.com (Andrew C. Esh) writes: >In article <1582@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >>In article <1991Apr08.171237.19978@shl.com> phil@shl.com (Phil Trubey) writes: >> >>>As fas as the hype being invented, every new installation that I've dealt >>>with (I work for a systems integrator) in the last year has been 10BaseT. >>>The *only* place where I would recommend coax is in a lab environment > >Only a lab environment? You must realize that some sites are more spread >out than what 10baseT can reach. Sure, it will work for one floor of a >small to medium sized building, but sheer physics drives you up into >ThinNet, ThickNet, and Fiber when the physical seperation between nodes is >great. 10baseT just doesn't go more than 400 ft. The company I work for is >called upon to design nets for buildings up to a quarter MILE on a side. >10baseT is great for an office area, but you need something else for >distance. >> You can daisy-chain MPR's at least up to four; at 300' each, you can safely install a 10BASE-T network whose far-to-near end components approach 1500' - you need to factor the Round-Trip-Bit-Delay using the constructs of the MPR and the cable length. Using higher quality wire (i.e., AT&T's 2061 cable), you can run even farther!\ To go one step farther, there *really* isn't a four-repeater RULE in 10BASE-T, and I have seen more than six MPR's strung out. However, for Distance (with a Capital-D), nothing beats FOIRL for that, unless you have the bucks for FDDI. -- +--------------------------------------------------------------------------+ | Carlo Milono: cmilono@netcom.apple.com or apple!netcom!cmilono | |"When a true genius appears in the world, you may know him by this sign, | |that the dunces are all in confederacy against him." - Jonathan Swift | +--------------------------------------------------------------------------+
arnold@synopsys.com (Arnold de Leon) (04/13/91)
In article <1991Apr11.044735.1221@ux1.cso.uiuc.edu> german@uxh.cso.uiuc.edu (Gregory German) writes: >Money is not the only issue here. For the most part I would claim >that it is asking for trouble to daisy-chain between offices and I >very rarely design a network that includes daisy-chains outside of >labs/classrooms. Your users are at the mercy of the guy in the next >office and everyone on that segement MUST be on the same network. >If you want to place one person on a separate network you have to >rewire. > We're in the process of moving into a new building which has 10 different subnets. UTP really wins for virtually being able to put any any office on any subnet. Our last installation was thin wire and it was exciting everytime someone rearrange their office, gee my bookcase would fit better if that little piece of cable wasn't there. It was even worse when the the network became too busy and and I had to find a way to take flat network and subnet/bridge it (It was impossible to group workstations to their servers). >I don't think anyone is really advocating pulling UTP. My main reason to >use it is to avoid having to pull cable into each office right next to >unused UTP pairs. If I have to install cable I would use thinnet, BUT >I would for the most part opt for homeruns for reliability, flexibility >and manageability. > We did pull UTP for the new installation. Aren't thin-wire ports more expensive that UTP ports on hubs? You seem to imply 1 port per user. >You have a valid point about colocated hosts being a problem with 10baseT. >The converters back to coax can work, but are in the $450 range and have >some limitations. I bought a couple of the converters for less than $300. > >My point is that there are other considerations than just cost and other >costs than just installation. IMHO there is and will continue to be a >place for both technologies. > Another thing I like about 10 base-T is the ability to monitor link status. Given the right hub you can tell if a host has been unplugged or powered down or up. You can potentially use this as a way to monitor access to your network. It would be more difficult for a user to add an unauthorized node. > Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) -- Arnold de Leon arnold@synopsys.com Synopsys Inc. (415) 962-5051 1098 Alta Ave. Mt. View, CA 94043
jbreeden@netcom.COM (John Breeden) (04/14/91)
In article <721@synopsys.COM> arnold@synopsys.com (Arnold de Leon) writes: > > Another thing I like about 10 base-T is the ability to >monitor link status. Given the right hub you can tell if a host >has been unplugged or powered down or up. You can potentially use this as >a way to monitor access to your network. It would be more difficult >for a user to add an unauthorized node. > I really don't think that LI by itself provides or adds any real "management" capability to 10baseT. 1. What good is it to know if a station is either unplugged OR powered down (or dead)? I need to know which it is, LI by itself doesn't tell me. 2. In terms of an "unauthorized" node on the net, it seem that refurs to an "unauthorized" mac address. Just unplug a "known" node and plug in a new device. LI won't know the difference (and LI "dis- apearing" momentarily dosn't help - LI has no way of telling WHY LI went away (see #1). 3. LI can't tell me how GOOD a connection I have, just that SOME type of connection exists. Real World: Bad TP wire run drops 10-20% of the packets sent - LI reports "all ok". LI IS a real handy function when INSTALLING wire/nodes (lets me know I have tranceiver to transceiver connectivity), but is pretty usless as a way to "manage" my network. Knowing that I don't have continuity is pretty useless until I know WHY. I've seen one MAJOR company that decided to alarm EVERY LI port to EVERY PC in their network. It lasted one day. At 5 o'clock their management system went crazy - everybody turned off their PCs and went home. -- John Robert Breeden, jbreeden@netcom.com, apple!netcom!jbreeden, ATTMAIL:!jbreeden ------------------------------------------------------------------- "The nice thing about standards is that you have so many to choose from. If you don't like any of them, you just wait for next year's model."
sph@logitek.co.uk (Stephen Hope) (04/15/91)
pat@hprnd.rose.hp.com (Pat Thaler) writes: >> > >> >- is there any advantage to "cascading" hubs (attaching the next one to a >> >port in the previous one - you lose two ports) as opposed to putting both >> >directly on a thin/thick backbone and using all ports for the UTP star? >> Lose two ports? If you are connecting two "out" ports together, you will >> have problems. To cascade, you need to connect an out port of the parent, >> to the AUI port of the child, with a 10BaseT transceiver. Without the tree >> heirarchy, the timing gets all crunched and one or both of the hubs will >> stop working until the problem is corrected. >There seems to be some confusion here. Perhaps it's because 1BASE5 >(1 mb/s twisted pair version of IEEE 802.3, usually called StarLAN) >had hubs it a special port to cascade to the next higher hub. In >any 10 mb/s IEEE 802.3 repeater, any port can be connected to another >repeater port. Repeater operation, timing, etc are the same for all >ports. >There is one thing you should check when connection repeaters together >by their twisted pair ports. Pins 1 & 2 are the TX pair for DTEs >and pins 3 & 5 are the Rx pair for DTEs. Inorder to allow wiring <Stuff deleted> Try pins 3 and 6 for the TX pair. >Pat Thaler Stephen Hope #include <std/disclaimer>
andrew@jhereg.osa.com (Andrew C. Esh) (04/16/91)
In article <1991Apr10.150801.2519@ux1.cso.uiuc.edu> german@uxh.cso.uiuc.edu (Gregory German) writes: > >Even though we don't use it in most cases I have also designed in the ability >to add intelligent monitoring tools to the the hubs. LANview with Cabletron >repeaters and an module with SNMP capability in the Plexcom chasis. I expect >to see increased use of these as tools become available. > >> - RL "Bob" Morgan >> Networking Systems >> Stanford >-- > Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) >US Mail: Univ of Illinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801 >Office: 129 Digital Computer Lab., Network Design Office Yes, and the David box, as well as the Xyplex Communications Server both have SNMP modules also. Real nice, since that means one's pink little butt and one's cushy office chair don't have to part company in order to figure out where the problem is. :-) Two thumbs up for SNMP. -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Mpls, MN 55416-1528 Punch down, turn around, do a little crimpin' (612) 525-0000 Punch down, turn around, plug it in and go ...
andrew@jhereg.osa.com (Andrew C. Esh) (04/16/91)
In article <1991Apr12.013427.24895@netcom.COM> cmilono@netcom.COM (Carlo Milono) writes: >In article <1991Apr9.221136.12326@jhereg.osa.com> andrew@jhereg.osa.com (Andrew C. Esh) writes: >>In article <1582@vtserf.cc.vt.edu> jcrowder@GroupW.cns.vt.edu (Jeff Crowder) writes: >>>In article <1991Apr08.171237.19978@shl.com> phil@shl.com (Phil Trubey) writes: >>> >>>>As fas as the hype being invented, every new installation that I've dealt >>>>with (I work for a systems integrator) in the last year has been 10BaseT. >>>>The *only* place where I would recommend coax is in a lab environment >> >>Only a lab environment? You must realize that some sites are more spread >>out than what 10baseT can reach. Sure, it will work for one floor of a >>small to medium sized building, but sheer physics drives you up into >>ThinNet, ThickNet, and Fiber when the physical seperation between nodes is >>great. 10baseT just doesn't go more than 400 ft. The company I work for is >>called upon to design nets for buildings up to a quarter MILE on a side. >>10baseT is great for an office area, but you need something else for >>distance. >>> >You can daisy-chain MPR's at least up to four; at 300' each, you can >safely install a 10BASE-T network whose far-to-near end components >approach 1500' - you need to factor the Round-Trip-Bit-Delay using >the constructs of the MPR and the cable length. Using higher quality >wire (i.e., AT&T's 2061 cable), you can run even farther!\ > >To go one step farther, there *really* isn't a four-repeater RULE in >10BASE-T, and I have seen more than six MPR's strung out. However, >for Distance (with a Capital-D), nothing beats FOIRL for that, unless >you have the bucks for FDDI. >-- >+--------------------------------------------------------------------------+ >| Carlo Milono: cmilono@netcom.apple.com or apple!netcom!cmilono | >|"When a true genius appears in the world, you may know him by this sign, | >|that the dunces are all in confederacy against him." - Jonathan Swift | >+--------------------------------------------------------------------------+ I'm sure that given any cable type, there is a combination of boxes that will make it run for miles. The problem with 10baseT for distance is that we are not trying to cover a square quarter mile of area, we have departments spread out in pockets all over that area. 10baseT in the department is fine. We need something else to cover the 500' runs in between. Up 'til now it's been Thinknet, but that's too small, and we needed to integrate voice and other stuff, so FDDI. 1/4 mile square manufacturing facilities are of a scale that FDDI becomes cost effective, mainly because nothing else provides the kind of bandwidth these folks wolf down. Yikes! "Imagine all the Routers, sending packets by. Imagine all Transceivers, pulling transmit high. Imagine all the users, writing to their files. You may say ... I'm a wiener, But I'm not the only one. When I'm through as a weiner, All the nets will be as one." (apologies, John Lennon) -- Andrew C. Esh andrew@osa.com Open Systems Architects, Inc. Mpls, MN 55416-1528 Punch down, turn around, do a little crimpin' (612) 525-0000 Punch down, turn around, plug it in and go ...
jbreeden@netcom.COM (John Breeden) (04/17/91)
In article <1991Apr15.214932.9635@jhereg.osa.com> andrew@jhereg.osa.com (Andrew C. Esh) writes: >In article <1991Apr10.150801.2519@ux1.cso.uiuc.edu> german@uxh.cso.uiuc.edu (Gregory German) writes: >> >>Even though we don't use it in most cases I have also designed in the ability >>to add intelligent monitoring tools to the the hubs. LANview with Cabletron >>repeaters and an module with SNMP capability in the Plexcom chasis. I expect >>to see increased use of these as tools become available. >> >>> - RL "Bob" Morgan >>> Networking Systems >>> Stanford >>-- >> Greg German (german@sonne.CSO.UIUC.EDU) (217-333-8293) >>US Mail: Univ of Illinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801 >>Office: 129 Digital Computer Lab., Network Design Office > >Yes, and the David box, as well as the Xyplex Communications Server both >have SNMP modules also. Real nice, since that means one's pink little butt >and one's cushy office chair don't have to part company in order to figure >out where the problem is. :-) Two thumbs up for SNMP. >-- >Andrew C. Esh andrew@osa.com >Open Systems Architects, Inc. >Mpls, MN 55416-1528 Punch down, turn around, do a little crimpin' >(612) 525-0000 Punch down, turn around, plug it in and go ... I've been paying around with a new hub - AT&T's "smart" hub - includes an SNMP agent (save the pink little but from work) and also has another in- teresting feature - security. These hubs know about every MAC address attached to each port. Press a button on the SNMP monitor and it draws a graphic map of all the hubs and each and every MAC address attached to it, I can build and alias file of the MAC addresses (x-ref'd to domain names) - but that's not the best part. If someone unplugs their box and moves, the map instantly shows his new location (wow! a REAL TIME physical map! - with addresses!). I can let the guy (or gal) stay at the new location or block him! - automatically! As a matter of fact, I can block ANY new MAC address from being added to the network and force a mac address to a specific port. But the most interesting feature is this - a frame sent to a specific mac address only shows up at the port that mac address is attached to. All other ports get the header with the data field blank! (filled with 1s & 0s) - you can't capture other people's traffic! only your own! (they're doing the filtering during the normal buffer copy in the repeater - so there is no impact on delay - it's within 10baseT spec). So now I've got an ethernet that acts sorta' like SNA. Users can only attach to a specific port and only traffic destined for that port appears. -- John Robert Breeden, jbreeden@netcom.com, apple!netcom!jbreeden, ATTMAIL:!jbreeden ------------------------------------------------------------------- "The nice thing about standards is that you have so many to choose from. If you don't like any of them, you just wait for next year's model."
spurgeon@sirius.cc.utexas.edu (Charles E. Spurgeon) (04/17/91)
In article <1991Apr16.182217.6151@netcom.COM> jbreeden@netcom.COM (John Breeden) writes: > >I've been paying around with a new hub - AT&T's "smart" hub - includes an >SNMP agent (save the pink little but from work) and also has another in- >teresting feature - security. > (details of hub operation deleted) > >But the most interesting feature is this - a frame sent to a specific >mac address only shows up at the port that mac address is attached to. All >other ports get the header with the data field blank! (filled with 1s & 0s) >- you can't capture other people's traffic! only your own! (they're doing the >filtering during the normal buffer copy in the repeater - so there is no >impact on delay - it's within 10baseT spec). > That's interesting. The old 802.3c-1988 repeater spec I have states in 9.5.5 ("Data Handling"): "The repeater unit, when presented a packet at any of its ports, shall pass the data frame of said packet intact and without modification, subtraction, or addition to all other ports connected with the repeater unit. The only exceptions to this rule are when contention exists among any of the ports or when the receive port is partitioned as defined in 9.9.6." I wonder if this repeater spec has been changed in the last few years? -ces Charles Spurgeon spurgeon@emx.utexas.edu --------------------------------------------------------------------------- Rule #1. Don't sweat the small stuff.
phil@shl.com (Phil Trubey) (04/18/91)
In article <1991Apr16.182217.6151@netcom.COM> jbreeden@netcom.COM (John Breeden) writes: >But the most interesting feature is this - a frame sent to a specific >mac address only shows up at the port that mac address is attached to. All >other ports get the header with the data field blank! (filled with 1s & 0s) >- you can't capture other people's traffic! only your own! (they're doing the >filtering during the normal buffer copy in the repeater - so there is no >impact on delay - it's within 10baseT spec). > >So now I've got an ethernet that acts sorta' like SNA. Users can only >attach to a specific port and only traffic destined for that port appears. Nifty feature! That should be worth something. How long has AT&T been selling these hubs? Are they their own, or are they reselling/OEMing? BTW, has anyone heard of a product that does per port *bridging* inside a 10BaseT hub instead of per port repeating? ie. Packets would be sent into the hub and the hub would switch it out to the sole destination port. With a fast enough bridging unit, you could up the bandwidth of your ethernet hub by an order or magnitude or two... -- Phil Trubey | Internet: phil@shl.com SHL Systemhouse Inc. | UUCP: ...!uunet!shl!phil 50 O'Connor St., Suite 501 | Phone: 613-236-6604 x667 Ottawa, Ontario, Canada | Fax: 613-236-2043
fitz@wang.com (Tom Fitzgerald) (04/18/91)
jbreeden@netcom.COM (John Breeden) writes: > I've been paying around with a new hub - AT&T's "smart" hub [...] > But the most interesting feature is this - a frame sent to a specific > mac address only shows up at the port that mac address is attached to. All > other ports get the header with the data field blank! (filled with 1s & 0s) > - you can't capture other people's traffic! only your own! You mean, you can't receive any packets until you've sent at least one packet out? Sounds like it could cause problems. If you clear the hub's memory while systems are attached to it, 2 machines that were talking to each other will be locked out of communication until they both send unsolicited packets and the hub figures out where they are. Protocols that depend on systems knowing each others' MAC addresses would have real problems at boot time. (Broadcast-oriented protocols should be ok). I'll bet you could get around this anyway, if you had a PC with a flexible enough interface card. Send out a lot of packets with different source addresses, i.e. all the mac addresses that you know of on the same ethernet that aren't attached to the same hub. The hub will think that all those nodes are on your port, and will give you the data. Of course, this will show up on your map. --- Tom Fitzgerald Wang Labs fitz@wang.com 1-508-967-5278 Lowell MA, USA ...!uunet!wang!fitz
jbreeden@netcom.COM (John Breeden) (04/18/91)
In article <1991Apr17.212748.7165@shl.com> phil@shl.com (Phil Trubey) writes: >In article <1991Apr16.182217.6151@netcom.COM> jbreeden@netcom.COM (John Breeden) writes: >>But the most interesting feature is this - a frame sent to a specific >>mac address only shows up at the port that mac address is attached to. All >>other ports get the header with the data field blank! (filled with 1s & 0s) >>- you can't capture other people's traffic! only your own! (they're doing the >>filtering during the normal buffer copy in the repeater - so there is no >>impact on delay - it's within 10baseT spec). >> >>So now I've got an ethernet that acts sorta' like SNA. Users can only >>attach to a specific port and only traffic destined for that port appears. > >Nifty feature! That should be worth something. How long has AT&T been >selling these hubs? Are they their own, or are they reselling/OEMing? > Since the first of the year (they've showed it at Interop and Netword). AT&T Makes it themselves. -- John Robert Breeden, jbreeden@netcom.com, apple!netcom!jbreeden, ATTMAIL:!jbreeden ------------------------------------------------------------------- "The nice thing about standards is that you have so many to choose from. If you don't like any of them, you just wait for next year's model."
ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) (04/20/91)
Again, thanks to all who replied (the topic evolved somewhat since I asked the question re. UTP vs. thinnet, so I hereby absolve myself of all responsibility :-) I have collected some of the responses (ca. 70K); I can send them on request. Eric -- Eric Behr, Illinois State University, Mathematics Department Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu
woody@ucscb.UCSC.EDU (Bill Woodcock) (04/20/91)
> jbreeden@netcom.COM (John Breeden) writes: > A frame sent to a specific mac address only > shows up at the port that mac address is > attached to. All other ports get the header > with the data field blank! (filled with 1s & > 0s) - you can't capture other people's > traffic! only your own! > phil@shl.com (Phil Trubey) writes: > Has anyone heard of a product that does per > port *bridging* inside a 10BaseT hub instead > of per port repeating? ie. Packets would be > sent into the hub and the hub would switch it > out to the sole destination port. With a fast > enough bridging unit, you could up the > bandwidth of your ethernet hub by an order or > magnitude or two... There is one AppleTalk hub out that does this, and more have been rumored. Although this is potentially neat, it makes a hash of the heirarchical models we're used to using when we think about our networks. It bottlenecks any transmissions to faster media, (FDDI, in this case, instead of Ethernet) and it renders all your expensive network troubleshooting and packet analysis utilities useless. If the hub's management software is, in and of itself, smart enough to give you those same features, (as is partially the case with Tribe Computing's AppleTalk hub) then you're partially off the hook, but how many Ethernet hardware comapnies would you trust to write good software? The only way to get around the gatewaying bottleneck is to add multiple gateways per hub, which greatly complicates addressing and troubleshooting, and starts costing _a lot_. -Bill Woodcock BMUG NetAdmin ________________________________________________________________________________ bill.woodcock.iv..woody@ucscb.ucsc.edu..2355.virginia.st..berkeley.ca.94709.1315
jal@acc.flint.umich.edu (John Lauro) (04/21/91)
In article <14740@darkstar.ucsc.edu> woody@ucscb.UCSC.EDU (Bill Woodcock) writes: > > phil@shl.com (Phil Trubey) writes: > > Has anyone heard of a product that does per > > port *bridging* inside a 10BaseT hub instead > > of per port repeating? ie. Packets would be > > sent into the hub and the hub would switch it > > out to the sole destination port. With a fast > > enough bridging unit, you could up the > > bandwidth of your ethernet hub by an order or > > magnitude or two... > > networks. It bottlenecks any transmissions to faster media, (FDDI, in > this case, instead of Ethernet) and it renders all your expensive > network troubleshooting and packet analysis utilities useless. If the > hub's management software is, in and of itself, smart enough to give > you those same features, (as is partially the case with Tribe > Computing's AppleTalk hub) then you're partially off the hook, but how > many Ethernet hardware comapnies would you trust to write good > software? The only way to get around the gatewaying bottleneck is to > add multiple gateways per hub, which greatly complicates addressing > and troubleshooting, and starts costing _a lot_. The hub would have to be extremely fast to bridge a large number of ports. Assuming worst case... Number of ports * the speed of ethernet. With 11 ports, you are talking about > FDDI speed. Migrating to FDDI should be easy, if it can be integrated right into the hub. The only problem is it wouldn't be cost effective at this point. The best price I seen on bridges that can handle full speed ethernet (filter rate of 29,600 packets/second) are about $2,300. I would hate to think of the cost for 50 port or 100 port hub with bridging on each port. (Could be better than FDDI if affordable. You could then run FDDI to your servers on the same hub, etc...) It would certainly eliminate any problems of collisions... The closest I've seen done is to have many small hubs, and bridge each of the hubs into a central hub.
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (04/22/91)
Be careful about enthusiasm for FDDI<-->ethernet bridges. One hint of the problems is that they stopped being called "transparent" and started being called "translucent bridges" in about 1989. An example of the problems can been seen by considering the implications of putting half of an NFS client/server pair the fiber and the other half on ethernet--the "bridge" must fragment UDP/IP packets to make things work. Also beware that some early FDDI bridges did not use RFC-1103, and so can talk only to each other. They "tunnelled" ethernet thru FDDI, instead of bridging between ethernet and FDDI. Vernon Schryver, vjs@sgi.com
phil@shl.com (Phil Trubey) (04/23/91)
In article <1991Apr21.021222.947@engin.umich.edu> jal@acc.flint.umich.edu (John Lauro) writes: >The hub would have to be extremely fast to bridge a large number of >ports. Assuming worst case... Number of ports * the speed of ethernet. >With 11 ports, you are talking about > FDDI speed. Not necessarily. You could have a single high speed backplane running at FDDI speeds, for instance, and use the backplane as the packet switching bus. Northern Telecom had (has?) a product out that did something like this. They had a proprietary network scheme where each port (out of up to 1048 ports I think) had a dedicated 2 Mbps. The internal switching bus ran at 40 Mbps, so your total 'network' throughput was limited to 40 Mbps. Come to think about it, cisco Systems is doing this now with their cBus architecture - they have a network bus running at something like 533 Mbps. -- Phil Trubey | Internet: phil@shl.com SHL Systemhouse Inc. | UUCP: ...!uunet!shl!phil 50 O'Connor St., Suite 501 | Phone: 613-236-6604 x667 Ottawa, Ontario, Canada | Fax: 613-236-2043
pat@hprnd.rose.hp.com (Pat Thaler) (04/26/91)
>There is one thing you should check when connection repeaters together >by their twisted pair ports. Pins 1 & 2 are the TX pair for DTEs >and pins 3 & 5 are the Rx pair for DTEs. Inorder to allow wiring <Stuff deleted> Try pins 3 and 6 for the TX pair. Oops, you are right that it should be 3 and 6. My short fingers sometimes have trouble hitting the right number key. But 3 and 6 are the Rx pair for the DTE, not TX. >Pat Thaler Stephen Hope Pat
cmilono@netcom.COM (Carlo Milono) (04/26/91)
In article <1991Apr21.021222.947@engin.umich.edu> jal@acc.flint.umich.edu (John Lauro) writes: >In article <14740@darkstar.ucsc.edu> woody@ucscb.UCSC.EDU (Bill Woodcock) writes: >> > phil@shl.com (Phil Trubey) writes: >> > Has anyone heard of a product that does per >> > port *bridging* inside a 10BaseT hub instead >> > of per port repeating? ie. Packets would be >> > sent into the hub and the hub would switch it >> > out to the sole destination port. With a fast >> > enough bridging unit, you could up the >> > bandwidth of your ethernet hub by an order or >> > magnitude or two... > ......[stuff deleted to protect innocent....] >The hub would have to be extremely fast to bridge a large number of >ports. Assuming worst case... Number of ports * the speed of ethernet. >With 11 ports, you are talking about > FDDI speed. Migrating to >FDDI should be easy, if it can be integrated right into the hub. The >only problem is it wouldn't be cost effective at this point. The best >price I seen on bridges that can handle full speed ethernet (filter rate >of 29,600 packets/second) are about $2,300 Ah, filtering is one thing, but forwarding is another and is always slower by quite a bit! >. I would hate to think of the cost for 50 port or 100 port hub with > bridging on each port. (Could be >better than FDDI if affordable. > You could then run FDDI to your servers on the same hub, etc...) > It would certainly eliminate any problems of collisions... And, when you bridge, you are going to bridge to all others? I suppose what you are asking for is a *star* arrangement, much like RS232-C, that supports *private* 10Mbps to a server from the client. No need for Carrier Sense, there would be no Multiple Access, and you would not need to Detect Collisions...= not 802.3 CSMA/CD! And how would you support multiple hosts? Link them via SONET/FDDI and 'telnet' through your 'local' host. > The closest I've seen done is to have many small hubs, >and bridge each of the hubs into a central hub. ...and you had better have a server on each hub or else you suffer the fate of the forwarding rate... -- +--------------------------------------------------------------------------+ | Carlo Milono: cmilono@netcom.apple.com or apple!netcom!cmilono | |"When a true genius appears in the world, you may know him by this sign, | |that the dunces are all in confederacy against him." - Jonathan Swift |
jal@acc.flint.umich.edu (John Lauro) (04/28/91)
In article <1991Apr26.032625.26585@netcom.COM> cmilono@netcom.COM (Carlo Milono) writes: > Ah, filtering is one thing, but forwarding is another and is >always slower by quite a bit! A high speed bridge should be able to forward at number of ports * ethernet speed / 2. Even if it can only forward at FDDI speed, it would be better than putting all those ports together. > And, when you bridge, you are going to bridge to all others? >I suppose what you are asking for is a *star* arrangement, much like >RS232-C, that supports *private* 10Mbps to a server from the client. No, a *private* 10Mbps to the hub. A server would have it's own private 10Mbps connection to the hub. A packet enters one port and only leaves one port (unless it is broadcast, or completely filtered out.) The next packet to enter, may be forwarded to a different port based on destination address... >No need for Carrier Sense, there would be no Multiple Access, and >you would not need to Detect Collisions...= not 802.3 CSMA/CD! And Unless of course you have multiple devices at one end of the tp. (IE a repeater from tp to thin net.) >how would you support multiple hosts? Link them via SONET/FDDI and >'telnet' through your 'local' host. The same way you do now. What's the problem? >> The closest I've seen done is to have many small hubs, >>and bridge each of the hubs into a central hub. > >...and you had better have a server on each hub or else you suffer the >fate of the forwarding rate... Not if your bridge has a decent forwarding rate. Have *you* actually tried it? The performance is better than having the servers compete with 500 computers for the same bandwidth. On the average it filters out more and reduces the number of collisions so that the overall speed is much greater.
ejbehr@rs6000.cmp.ilstu.edu (Eric Behr) (05/01/91)
The discussion on 10Base-T and its advantages seems to be a hot topic: so far I've had about 30 requests for copies of the summary. -- If you missed it the first time around, I make the offer again. I can e-mail the summary of opinions I've collected (ca. 70K, in 3 parts); -- If you did request it, but didn't receive it, please e-mail me again. There were a few days when our mailer was acting up, so I may have lost some letters. -- Eric Behr, Illinois State University, Mathematics Department Internet: ejbehr@rs6000.cmp.ilstu.edu Bitnet: ebehr@ilstu