why is Token ring expensive than ethernet

When TR was a 4Mb/sec ring and Ethernet was a 10Mb/sec shared bus there were all sorts of studies that claimed that "colisions" would make Ethernet unpredictable above some threshold. In a corporate environment reliability and maintainability have a high priority, as it should be. When used in an all-IBM environmant there were system management and diagnstics tools that Ethernet and IP didn't get until managed hubs came into use.

When Ethernet was a stiff orange coax cable as thick as your pinky TR cable didn't look that bad.

Only a few companies made TR cards and becasue of interoperability problems with PC brands a company tended to buy one brand of card and stick with it. There wasn't lots of competition and stuff was priced for huge businesses, who look for stability and vendor support.

Just a matter of sales volume. No point sticking with Token Ring though, is there?

In the early days of LANs, there were many different flavors, each with its zealous and vociferous proponents. I suppose that was also true in the early days of the automobile, with steam, electric, and gasoline powered cars vying for public acceptance. After that initial period, winners emerge and the others become historical oddities.

Happened with wired LANs too.

Bert

In article , sharkie wrote: : I am new to networks. I have read that Token rings are more :expensive than ethernet. They seems to be using same amount of :cabling, don't they ?, is it expensive to build the NICs for Token :ring ?, is it management cost that is higher in the case of Token ring

I have never used or priced Token Ring, but I can offer some hypotheses.

In modern switched Ethernet networks, the communication between different computers can often proceed in parallel. With 10BaseT,

100BaseT, and 100BaseTX, each end-node -must- be connected to a switch or router [more exactly, each segment must consist of exactly two devices.] With full duplex, that implies no collisions on the segments, and thus no need for elaborate mechanisms that are designed to prevent collisions. There are still flow control issues, but they are taken care of by buffering [and "atomic locks"] in the switches, with packets dropped only when the switch buffers get full. One can chain together additional simple switches and routers and additional devices with little concern about the effect upon other devices.

Token ring, on the other hand, expects looped segments, dual cables to each device, and that the NICs fail to pass-through [so if a device dies, the next device in the chain still gets the signal.] One cannot simply hook up more devices and leave ports unconnected at will: one must maintain the loop. Which is fine if your setup is quite static, but is not condusive to rapidly changing environments in which users expect to be able to just plug in or unplug anything that has a network interface. Token ring bridges certainly do exist, but if you want to be able to get to the easy plug-in convenience of Ethernet, you end up having to go to the star "exactly two devices per segment" topology, in which case it becomes mostly redundant to bother with the token.

This situation leads to a greater demand for switched ethernet than for token ring, so companies put more resources into mass production of ethernet devices, get "economies of scale", with the result that the prices keep falling on ethernet devices and manufacturers have little incentive to make cheap token ring.

see the other posts for some perspectives based on scale and market share.

the other issue is that token ring is a lot more complex - complicated NIC cards (which must have their own processor to handle ring management), complex design rules amd / or complex hubs.

And source routing (used in T/R bridges and switches) is complex, and needs NIC type managment, and is difficult to do in a hardware accelerated chip.

Source routing in a resilient networks also tends to generate multiple copies of a broadcast, so it was diffcult to build large resilient T/R systems - and they had to be planned carefully, or they didnt work reliably.

So - slower (although i did work on some IBM 100 Mbps T/R) Shared (but switched / full duplex T/R eventually appeared) complex (and the support complications probably helped to kill it off) expensive

and finally - it was really tied to IBM - there was a set of IEEE standards, but wherever IBM went a different way, so did all the other manufacturers to maintain compatibility with the dominant manufacturer.

A bit irrelevant now as i dont know of anyone making new T/R cards, hubs etc. (well - maybe Cisco et al will sell you an interface for a router, or IBM a NIC card for a mainframe....)

it used to run out 2 to 3 times the price - but there wasnt any T/R equivalent to cheap Taiwanese Enet NICs, so that was only comparing leading (=expensive) brands.

The real reason for the cost difference was very simple--IBM controlled the technology and charged a premium for the license. Now it's expensive because there is only one vendor doing a relatively tiny sales volume while Ethernet has become a mass-market consumer product.

As for the percentage, well, the cheapest Token Ring NIC on the market goes for about $135. New Intel 100TX NICs go for about $25 or white box Taiwanese for $4. The cheapest Token Ring switch on the market is a managed 24 port for $2650. New HP J4813As go for about $435, or you can go to a D-Link with roughly equivalent functionality but not quality for about

360. Or if you want to go to an unmanaged switch you can go as low as $70 for 24 ports.

So, lowest price for NICs and switches for 24 nodes of 16 Mb/sec Token Ring would be about $5890 vs 166 for 100 Mb/sec Ethernet. For equivalent quality and functionality 100 Mb/sec Ethernet would cost $1035. To that add the cabling cost--at $100/drop (a conservative number for the most part) that would be $2400 for any of them. Then there's delivery and installation of the hardware--for a network that size that's pretty much plug and play so call it a day to get everything set up and tested--that's another $800 or so depending on your labor rates and overhead. So, minimum cost for Token Ring would be about $9090, an equivalent but much faster Ethernet about $4235, and a minimum cost Ethernet about $3560. So 16 Mb/sec Token Ring, installed, would go 2-3 times more than 100 Mb/sec Ethernet at this time.

Of course that's all for new--if you're willing to go to used equipment Token Ring hardware gets dirt cheap--if you search hard enough you may find someone willing to pay you to haul it away.

No, they used more expensive cabling - IBM Type-1 cabling is rather hard to work with, and the hermaphroditic connectors take a whole lot more work to install than an RJ45.

Yes. Each token ring nic needs its own microprocessor, where an Ethernet nic has just a few fifos.

yes. Tracking down a problem in a token ring requires young guys with good ears to run over to a closet and find the connection that's clicking. The experienced guys who spent too much time in raised floor rooms would lose their hearing from the incessant AC noise and not be able to hear the tell-tale click of a machine beaconing.

Let's use 96 nodes, since ethernet devices tend to come in 12's and token ring devices in 8's. For a high-end gigabit deployment today, I can drop two Cisco gigabit switches in for about $20K total. Gig nics are about $70 each, so another $7000 there, and wire is about $300/drop installed in a corporate environment, so $57K.

Token ring nics are about $450 each. You can get an 8-port 16-Meg TR switch for about $10K, and the wire is more expensive and harder to pull and terminate - we'll call it $500 per drop. So, that's $45K in nics, $120K in switches, and $50K in wire - $215K, or almost 4 times the cost for 1/60th the bandwidth.

I put in a switched token ring environment (with ATM backbone) for a state government agency about 6 years ago, so these costs are pretty close. They ripped it all out and switched to ethernet after about 3 years.

When TR came out Ethernet was the original think orange coax with vampire taps. The IBM design looked very good for wiring an office in comparison.

The original 100 node/500 meter limit looked fine in 1980. In the original design I don't think anyone gave any thought as to how to link thinknet segments together to extend them. For a Data Center with raised floors, ethernet was fine.

Token ring never got faster than 16Mb/sec as far as I know, and many people though that 10Mb/sec ethernet was as fast or faster, most of the time. As soon as 100Mb ethernet with managed hubs hit the street TR was dead for new projects, except for diehard IBM shops.

The cost of the NIC is small potatos in the budget for a corporation wiring a facility. Wiring costs are much higher, but even they get amortized over years. Reliability and Time required to loacte and fix a problem very important and corporations will pay for it. TR was ahead of Ethernet in this regard until managed UTP/STP hubs were developed.

What did he say ??? :-)

IME The SNA guys could identify failing equipment without getting out of the chair in the warm, quit, office.

Token Ring runs fine on CAT5 and many Token Ring NICs and access units have RJ-45 connectors.

That's odd, I've never seen a microprocessor on a Token Ring board.

Doesn't work with anything but the hard-relay MAUs. Switches don't click. The drivers usually report beaconing on any reasonably modern machine.

Why does CAT5 for Token Ring cost more per drop than CAT5 for Ethernet?

Check the Madge site and Froogle your prices.

IBM was shipping 100 Mb/sec Token Ring ages ago. The trouble with it was that the switches cost vastly more than for Ethernet. And they still cost vastly more than for gigabit Ethernet, and no doubt if they're still available when 10 gig over copper goes mass-market they'll cost vastly more than that as well. One thing IBM never got, even after the PC hit, was the effect of competition on cost and performance--for Token Ring to continue to be a player it would have to go mass-market--there's no way that anybody could continue to justify it over something that was both much cheaper and much faster.

The cost of an Ethernet NIC. The cost of a Token Ring NIC and switch port is considerably more than the cost of the drop in most cases.

His experience with Token Ring is clearly more outdated than mine, and mine's pretty outdated.

Madge sells 100MBit TR components and IIRC there is at a specification for 1000MBit TR but no has implemented it (and probably no one ever will).

Jens

Madge sells 100MBit TR components and IIRC there is a specification for 1000MBit TR (802.5v?) but no has implemented it (and probably no one ever will).

Jens

We actually bought four of them. It had a GigE bridge built in to interface with 100Mbs TR. We needed the additional BW until we could do a POTPIE project. PullOutTR and PutInEthernet.

Some IBM routers supported 100M T/R (and i think there was a card for a mainframe front end - ro you could use the router for that).

once you get to 100M, T/R used the same cable interfaces as Ethernet,

BTW 100M T/R was switched, full duplex, so no Tokens, no Ring, no access mechanism, no beacons. It almost looked like Ethernet with large frames and some priority signalling......

IBM gave up on networking hardware (incl token ring switches, hubs and routers) in a technology / patents for money deal a few years back

they formally announced that they would suggest anyone wanting these types of hardware use Cisco...(i was working of an IBM network reseller at the time, and this killed that part of our business - apart from 2nd hand gear for companies who couldnt change quickly....).

Some bits and pieces probably survive - i havent checked for a while.

See

if you are interested.

MAUs got replaced with better electronic hub designs a long time ago - i think Synoptics and Proteon were a couple of the early companies - and both got eaten by others over a decade ago.

Hi Sharkie,

token ring has become a standard like Ethernet and other shared media systems. They are all defined under IEEE 802.*. Token RING (TR) was an invention of IBM and was offered with different speeds, such as 4 mbps,

16 mbps and there was also standardized a FastTR with 100 mbps. Historically the big brother technique of FDDI, which was standardized by ANSI has overtaken the basic idea of FDDI.

It has been working fine and relyable for a long time. It fascinated due to its predictability and accuracy in function. It was fairly managable and could be prioritize data in a very early time of development.

The engineers might think they have found THE solution for the market. But since this market is free and only the one with the best control concerning expenses and wins survives, this market does not evaluate genious solutions but costs.

The better product is the foe of the good product. The better product is the product that works as well as the good ane but develops lower costs. This is a very simple logic.

The prodoction numbers of Ethernet equipment rose but not one for TR. Due to company internal processes relying on TR, due to hardware that relies on TR that can not be given up so easyly, it still is present on the market. But it only fills a small gap.

Due to minimal presentation on the market, the big companies deny to support these old technologies, since they are bound to cut off products that produce costs over wins. This is a self regulating mechanism.

To cover the costs the bills these companies write during last time period of support on the market make every one responsible for equipment and costs freeze. This is a good motivation to get ridd of that technology as fast as possible.

The future belongs to Ethernet. Just see the history

1973 invented with a speed of 2.94 mbps 1980 standardized by IEEE 10.00 mbps 1992 (about) 100.00 mbps 1999 (about) 1000.00 mbps 2001 10000.00 mbps ???? 40000.00 mbps

Sharkie, I hope I could give you an idea of the market function of products. Network equipment is a product like any other. Mean While it is used in the production chain of allmost all products on the market, such having a burden influence on prices per peace.

Good luck to you and continue asking questions. There is a lot of expertise out here.

He> Hi,

You can still get token ring interfaces for Nortel/Bay/wellfleet routers. Madge (which had a very large percentage of the market share for token ring in fortune 500 sites) still makes PCI token ring cards for PC's.

99% of the installations I worked with used cat5 UTP cable with rj45's for tokenring. Most token ring boards manufactured in th 90's in fact needed a special adapter to use them with type1 cable.

It got up to 100mb. You can still buy new 100mb token ring cards (backward compatible with 4/16) to this day.

It's not dead yet. It's still around. I know out local hospital, (which is internationally known) still uses token ring.

It is not possible to construct a Token Ring network with performance equivalent to the previous generation of Ethernet at _any_ price, let alone the _current_ generation just coming to market.

The only 100 Mb/sec Token Ring NIC still in production costs approximately ten times as much as a cheap gigabit Ethernet NIC or about 3 times as much as a fancy one. The price of switches is astronomical.

The reason Token Ring costs more is that it is based on technology which must be licensed from IBM. Further, since it is now a low-volume niche product the economies of scale that have brought down the price of far more complex Ethernet hardware do not apply to its production.

In any case, Token Ring is of mostly academic interest at this point--there are still Token Ring shops out there, but they are few and their numbers are decreasing.