why is Token ring expensive than ethernet

(snip)

Maybe the government should file an anti-trust suit against the ethernet monopoly. Having competition is supposed to be good, isn't it? :}

-- glen

Not mentioned in this entire thread was that in the early days when TR and Enet were starting, IBM wasn't concerned with small shops. They were a bother. At about 91 or so when a 10baseT port on a hub got down to $100 a technically adept person could network a few computers and printers on their own. With TR IBM just ignored you. As the mass of small businesses and even departments within larger businesses started buying Enet because they "could" the economics started to swing.

No. IBM DID go after smaller shops - that is what the IBM baseband and IBM broadband network adapters were for. Not that ayone actually BOUGHT them. Small shops at the time did not use ethernet either - THAT was to expensive (an ISA 10base2 card would set you back over $750.00, not to mention the cost of stringing coax all over) Most small shops (under 100 or so people) tended to use arcnet, and SMC was the brand of choice.

They may have had products but they wouldn't talk to you about it unless you went to a reseller. In thinking back, about 92 to 94 I started putting Enet into location with 2 to 10 computers and TR wasn't even a consideration.

IMHO TR and micro channel were the last of the major attempts at IBM to do things the "old" way. We IBM sold their network division to Cisco about 10 years ago, TR went to life support. And MC just died off from lack of buyers.

In the 80s I was with a small software company. We sold mini-computers into the insurance agency market. IBM continually pressed us to switch. Along with our corporate partners. But the net of any switch would have been to double our net installed price into the offices where we were selling. And IBM just couldn't understand why that was a problem. In their opinion their name would have justified the price! And this doesn't even deal with the issue that switching to the IBM solutions would have cost the users features and productivity.

And yes, the systems and company where I worked went away also. Taken out by PCs. :)

Yes. And they really did not give the resellers ample training to support the stuff (I know firsthand, since I worked for an IBM reseller at the time)

TR was still around in force then. Madge was makeing most of the boards being installed at the time, and they sold pretty well. 92-94 - Chemical Bank (which later absorbed Chase Manhattan bank and took it's name was nearly 100% token ring (they strated replacing it around 95 or so) It's STILL around to this day, but only as legacy. A world famous hospital down the street from me still uses PS/2 micrchannel boxes running 3270 emulation over 4mbps token on type1 cable!

Novell used to SWEAR token ring at 16mbps was faster than shared ethernet at 100mbps because collisions would slow the ethernet down. They's teach that in all their official certification classes.

Just outta curiousity, what brand mini's? Dec? Prime? Wang?

That and larger packets.

I used to support Data General, DEC, Prime and Collins. The Collins systems were actually MIL spec versions of IBM gear, made under license.

Wang. But not the VS stuff. The 2200s. And if you know anything about those besides they existed, you're a rare person. :)

IBM was pushing us to use a funky workstation thing that could run up to

Big IBM shops used SNA and could diagnose anything, anywhere remotely before 1990. For a large company that's important. ThinWIre was certainly horrible as an infrastructure cabling scheme. There was good reason to avoid ethernet to the desktop until UTP and managed hubs came in. When I left the IBM/SNA world in 1993 ethernet was unmanagable by IBM standards.

Back then a cable scanner was the one tool that everybody running Ethernet needed and couldn't afford :(

I still have the Fluke TDR that I talked my boss into buying when we picked up responsibility for support of a small business that had a ThinWire LAN that been installed by a TV cable installer that used push-on BNC connectors and splices all over the place without using the crimp ring.

I walked away from a serial line datascope in 1992. ($35,000 suitcase unit with SNA, Decnet, and TCP protocols because it looked like modems were dead, and it was _HEAVY_ The web and home internet took off after that and I could have used it.

But the 2 to 50 computer networks could never afford those tools or the staff to monitor them. And thats the market that fueled the growth of ethernet while token ring was growing only at the rate of installed IBM terminals and PCs pretending to be 327x and such terminals.

There was one network that was having intermittent problems that I couldn't figure out. Finally broke down and bought a TDR. Took me about ten minutes to find that it was configured as a star of buses with three terminators and a tee connecter where a repeater belonged (the repeater was sitting there but somebody had decided to replace it with a tee connector). If I hadn't gone after it with the TDR I'd have never found either though--you had to _know_ that there was something there before you'd do enough digging to actually find it--it was well-hidden.

Very true. Even today, we are talking about "how do we show true end to end user experience..."

So everyone carried a terminator in their tool box! :) Binary search works pretty well :)

(snip)

Well, token passing does take some time, especially on a large ring. TR does have an advantage in that arbitration time increases as the ring gets bigger, without a fixed maximum round trip time.

Ethernet arbitration (collision resolution) is fairly fast, especially on a small (in distance) network.

TR hardware, especially including the logic to regenerate lost tokens, is significantly more complex. With enough economy of scale more complex isn't always bad, but TR has not done as well in economy of scale as ethernet.

Otherwise, used TR stuff is probably cheaper than used ethernet.

-- glen

One result of the above is that it is more fair in letting all hosts have an equal chance to use the wire. Though it does that independent of the frame size, so hosts with larger frames get a larger fraction of the bandwidth.

If only one host is transmitting it takes one token round trip time between frames. That is much less than 100%. If two are (more usual with ACKs coming back) then it is one round trip time per pair. Assuming no token loss, that is.

On the other hand, the ethernet capture effect gives more bandwidth to a host that can keep the wire busy. That may increase the usage for a situation involving heavy traffic from a small number of hosts.

How close will depend on ring size and frame size.

For ethernet, it depends on the (wire) distance between hosts actually transmitting and frame size. Jumbo frames can be used to increase the utilization in some cases.

-- glen

Token ring networks are expensive but they allow 100% usage of the bandwidth the network is capable of at all times, the reson being that only 1 person can send at a time and the bandwidth is dedicated to them untill they are done sending.

In article , Sean Pinegar wrote: :Token ring networks are expensive but they allow 100% usage of the bandwidth :the network is capable of at all times, the reson being that only 1 person :can send at a time and the bandwidth is dedicated to them untill they are :done sending.

I was never a student of Token Ring, but I know that what you are saying is debatible.

In order to be the transmitter, the host must have received the Token from downstream. When a host receives the Token and it has nothing to send, it allows the Token to continue onward, but if it does have something to send then it grabs the Token, starts sending, and doesn't emit a Token until afterwards (if I understand correctly.)

Suppose that A and B are communicating, and that they are adjacent in the ring and that there are other members of the ring (say C..Z). 'A' transmits, lets go of the token, B grabs the token, replies to A. Now, B releases the token -- and it has to flow through C, D, E, all the way to Z before finding its way back to A. During that time, the only signal on the wires is the token itself.

There are some who would argue that the time during which the token is in transit is not "usage of the bandwidth that the network is capable of at all times", and thus that Token Ring does not allow "100% usage". *Close* to 100%, perhaps, but not 100%.

It depends on what your definition is of "usage of bandwidth" and of what "the network is capable of".

Keep in mind that CSMA ethernet does not -disallow- complete occupancy of the wire with a combination of preamble, packets, and minimal intra-frame gaps (IFG): it just becomes -unlikely- as the number of devices that want to talk increases. Some would say that the IFG is wasted bandwidth; others would say that the IFG is simply a necessary part of "what the network is capable of".