Slashdot Mirror
Does Anyone Still Use Token Ring?
blanchae asks: "Does anyone still use Token Ring, or is it dead? I remember hearing about 100 mbps TR a few years ago but nothing since. I remember that the strong point of TR over Ethernet was the QOS and the consistent response time. Does the banking community still use TR?"
There is clearly a lot of research going on, with results published, as in Raja's Optimal bandwidth utilization in wireless token ring networks released earlier this year. However, 1998 was the last big year for user's guides, which indicates that this technology has long since fallen from the mainstream and now survives only in academia.
About 5 years ago, I worked for a trucking company that was very heavily invested in token ring and would not consider switching to ethernet, no matter how compelling the argument was at that time. I can only imagine it's harder to justify staying on TR now, but it can't be cheap.
Its only advantage was that it could run at much higher utilization than ethernet without your network choking - we would see times where a ring would be running at 75% and that was no problem.
However, it was a real problem from a financial and operational standpoint. When we bought new PC's, we would rip out the ethernet cards and install Olicom TR cards we paid $180 each for - we got a good deal because we bought hundreds of them. Server-class cards were more - a lot more.
And we did get the 100Mb token ring switches, which was truly one of the more absurd things I have ever seen IT money spent on. I still don't have a clear idea how this was a good thing: you got a 100Mb token ring switch, which would create a ring on each port. Then you could plug exactly one device into each port, as long as it had a 100Mb token ring adapter. This was 5 years ago, and I remember that per port, it was price-competitive with Gig-E fiber.
Then there are the usual entertaining issues with drivers and growing the network. Need an extra PC at your desk? You can't just plug a hub in and go - you have to pull another cable from the wiring closet. You need certified drivers for your Windows cluster? How about a touch-screen network device for your truck terminals? A firewall? A NAS? No, you can't have any of those.
I know there are plenty of people who will swear by TR. You'll find the evolved version of this technology in FDDI rings - and it makes a lot of sense and works very well in that application. But as a LAN for your company, it sucks ass...technically, the concept is sound but nobody is developing it further and it takes a lot more specialized knowledge and maintenance overhead than ethernet. And every year that goes by makes it much more expensive to keep it than to switch to ethernet.
I turned down a job 3 years ago at a place that was still running TR - a mid-sized retail chain. They said they were starting to look in to ethernet, but were happy with their token rings. That was the deciding factor for me to keep looking...At this point, a company that isn't actively working to replace TR with something else has some serious management issues and I would wonder what else was lying inder the surface.
So if you can find a few cards and a MAU somewhere, experiment with it at home. But avoid it like the plague in a business setting. That's just my $.02 anyway.
I experienced P2P token ring back in college. Here's how it worked: a group of peers arranged in a circular manner would pass around a named pipe. Each peer would hit the pipe, a process known as token. After a while, the pipe would be cached, and a designated peer reloaded the pipe.
I see token ring still in use in bank branches, main bank data processing centres, and some insurance companies. NATO is rumoured to have a bunch of legacy systems on TR. On the PC side, its mostly old ISA cards, and the 486-PII era machines which still have some crappy 32x0 emulator running in fullscreen mode on OS/2. On the the mainframe side, there are still old IBM 3080s+3090s, system 36/37/38s and many C390s around. Be afraid, be very afraid.
One of the side effects of some companies locked into dino^H^H^H^Hlegac^H^H^H^Htime tested solutions, is that they have to pay whatever it takes for dino^H^H^H^Hexperienced old-fa^H^Htimers to come in and fix the fsckups caused by young ignoramuses not having any knowledge of TR. My going rate right now is EUR400/hour, with a minimum of an 8 hour payment up front before I even set foot on the premises, and I still get called out about 3 times per year. get off my lawn...
Cisco must still have TR, I met a dejected CCIE candidate who told me he paid many thousands of euros for a one week CCIE-mill course, which took him from windoze point and click to supposedly a CCIE, only to have half his stack be wired with TR which the fly-by-night company had never heard of. Clearly the CCIE proctors have some tricks up their sleeves when they detect a candidate who has all the answers but none of the experience.
the AC
As well, my cisco study kit still has some 2513s and AGS+s and a box of TR cables (hermaphrodite and RJ45), ISA cards, and some 8228s. I haven't touched any of it in at least 5 years
Hemos is like...sci-fi fans;he thinks technology is cool, but he hasn't bothered to understand the science it's based on
I work for a bank... a LARGE bank.
I know we have one small location that still has TR because the site has been on the chopping block for 4 years.
(It's finally closing this year.)
I know we stopped installing it in new locations about 10 years ago in favor of Ethernet. My site (and most of the rest of the bank) was upgraded from TR to Ethernet(100Mb) about 5 years ago.
Banks and any other large companies are going to stick to industry standards in order to reduce costs and complexity. I know we've had a hell of a time finding replacement hardware for the switching/routing equipment in that last TR location. My point is, why should a large company build a custom LAN network when the cheaper, easier technology will do just fine. e.g. We would have to disable the ethernet adapter in the Dell workstations we use and install TR cards. I have a laptop...I'd have to find a PCMCIA TR card. This is exactly the type of BS that large companies don't want to deal with.
Here's the real reason TR is dead: QOS was only an issue with Ethernet when you had people using hubs. Now that massive switches are the norm, it isn't an issue since each user can run in full duplex. If you're on a hub, you're sharing bandwidth. If you're on a switch, you've got 100Mb all to yourself. (Unlike a hub, the switch can buffer the frames if the destination port is busy.) In addition, you can run in duplex which means your ethernet card can send and receive at the same time. If your office is using a switch, it's your WAN connections you have to worry about, not your LAN.
And thats just for the cube farm. For the server room we have either dual 100Mb or dual 1000Mb connections to multiple backbones (more for redundancy than bandwidth.) There are also dedicated fiber going to SANS drives.
The computer in my cube is piggy-backed onto a Cisco IP phone, which all goes to a single 100Mb switch port. I have never had a problem with it.
Token Ring is DEAD. DEAD. DEAD.
$7.95/mo, 200 GB disk, 2TBxfer, MySQL, PHP, RoR.
The big deal with token ring was that the network would remain stable under 100% load. Classic 10mbps ethernet with hubs would start experiencing trouble around 60% load and collapse by the time load reached 90%. If you had a big, flat network it just plain wouldn't work.
Look at why: With token ring, only the card holding the token could transmit. Everybody else had to wait for the token. So each station would empty its transmit queue and then pass the token on to the next station. On ethernet, a station would send a packet whenever and if another station sent a packet at about the same time they'd collide. Every station observing the collision would assert a collision signal and after the collision signal cleared the two stations that transmitted would wait a random period of time and then retransmit. That's oversimplifying a bit but more or less correct.
So, token ring was much more stable in a large LAN with a high probability of multiple stations having outbound traffic ready at the same time.
Now, along comes 100baseTX on cat5, the end of coaxial ethernet and the proliferation of $50 switches. When you're plugged in to a switch there are only two devices in the collision domain: you and the switch. So, lots less collisions. When you're in full duplex mode (as you generally are), collisions are impossible since by definition both sides are allowed to transmit at the same time. Now your ethernet network remains stable at 100% utilization. And if the nic in the PC burns out, the rest of the network doesn't care.
Token ring is very sensitive to malfunctioning nics. A malfunctioning nic may drop the token, that is it may receive the token and then fail to transmit it to the next nic. That kills a token ring network dead until the admin wanders around with an analyzer and figures out which PC is at fault.
Suddenly the tables were turned. Token ring was an administrative headache and expensive to boot. Ethernet was simple, cheap and worked just as well.
Token ring died out except as an academic curiosity -- an interesting early answer to a problem that was eventually solved another way.
Moderating "-1, Disagree" is simple censorship. Have the guts to post your opinion.
I've had one customer for whom token ring on Shielded Twisted Pair wiring was the right choice even after Cat5 Ethernet cards were cheap - they had lots of Big Electrical Equipment, and the alternative would have been to do fiber, which was cost-prohibitive back then, plus they didn't really need high data rates.
Performance differences weren't really all that significant for the different technologies, except for obvious base-rate differences (100 Mbps >> 16 Mbps > 10 Mbps > 4 Mbps.) Even if they were, Full-Duplex Ethernet (which is pretty much universal these days if you use switches instead of hubs) doesn't have the same issues that half-duplex does.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Myself and my wife work for IBM. One of my wife's first jobs at IBM was writing Token Ring drivers for early iterations of the NDIS interface. She had to write all the code on a 3270 terminal connected to a mainframe and cross-compile to the PC because the PC's couldn't handle the code. I joined the company two months before the Networking Hardware Division (which made Token Ring cards, ATM switches, Ethernet switches, mainframe communication devices, and Multiprotocol Routers) was paid $2B by Cisco to go out of business.
The Token Ring products were withdrawn from marketing a couple of years ago, so no more MAU's and Concentrators or NICs can be purchased, at least not from IBM. However, the products are still supported, and not uncommon in mainframe installations.
At IBM we finished the Ethernet migration a couple of years ago. The thing that struck me the most about the migration was how converting from 14Mbps TR cable to 100Mbps Ethernet cable involved nothing more than inserting an adapter cube into the connector on each end of the building cabling. One of the primary features of the "IBM Cabling System" was that it could be adapted to many different cable types by just using adapters; coax, twinax, UTP, etc. To accomplish this feat, it was actually shielded, as opposed to unshielded CAT3/5, etc. This made it hideously over-specc'd for the original common use of TR. The cabling was designed so you could run it past just about anything and not have to worry about interference, cross-talk, etc. You could even get cable that had some UTP pairs stuffed between the shielding and the sheath so you could run your phone and data cabling using the same cable run.
The drawback was that the cabling was bulky, expensive, and difficult to work with.
Making cable that will actually work at over six times it's origninal intended speed while being more than a bit difficult to work with is an interesting example of Enterprise-quality engineering philosophy at IBM from the '80s.
SirWired
(..to the tune of "Particle Man")
Token Ring LAN, Token Ring LAN
Doing the things a token ring can
How does it work?
It's not important
Token Ring LAN
Is it a drag or is it a waste?
When it's installed
Does it get replaced?
Or does that admin get axed instead?
Nobody cares
Token Ring LAN
Ethernet LAN, Ethernet LAN
Ethernet LAN hates Token Ring LAN
They have a fight
Ethernet wins
Ethernet LAN
Internet WAN, Internet WAN
Size of the entire Internet, man
Usually kind to the smaller LAN
Internet WAN
It's got a link with PPP band,
A T1 band, and an OC3 band
And when they're together it's a happy LAN
Powerful WAN, Internet WAN
Workgroups LAN, Workgroups LAN
Formerly known as MS LANMAN
Lives its life in a garbage can
Workgroups LAN
Is it depressed or is it a mess?
Does it feel totally worthless?
Who came up with Workgroups LAN?
Degraded LAN, Workgroups LAN
Ethernet LAN, Ethernet LAN
Ethernet LAN hates Token Ring LAN
They have a fight
Ethernet wins
Ethernet LAN
In theory, Ethernet on coax should be stable under heavy load. But in the late 1980s and early 1990s, it wasn't, due to defective design of some widely used interface chips. Here's the actual story. See this note by Wes Irish at Xerox PARC
The worst device was the SEEQ 8003 chip, found in some Cisco and SGI devices. Due to an error in the design of its hardware state machine, it would turn on its transmitter for a few nanoseconds in the middle of an interframe gap. This noise caused other machines on the LAN to restart their interframe gap timers and ignore the next packet, if it followed closely enough. This happened even if the SEEQ chip was neither the sender or the receiver of the packets involved. So as soon as you plugged one of these things into a LAN, throughput went down, even if it wasn't doing anything. A network analyzer wouldn't even see the false collision; this was at too low a level.
This was tough to find. Wes Irish worked on the problem by arranging for both ends of Xerox PARC's main coax LAN to terminate in one office. Then he hooked up a LeCroy digital oscilloscope to both ends. Then he tapped into a machine with an Ethernet controller to bring out a signal when the problem was detected and trigger the oscilloscope. Then, when the problem occured, he had a copy of the entire packet as an analog waveform stored in the scope. This could then be printed with a thermal printer and gone over by hand.
Because he had the same signal from both ends of the wire, the wierd SEEQ interference mentioned above appeared time-shifted due to speed of light lag, making it clear that the interference was from a different node than the one that was supposed to be sending. You could measure the time shift and figure out from where on the cable the noise was being inserted. Which he did.
It took some convincing to get manufacturers to admit there was a problem. It helped that Wes was at Xerox PARC, where Ethernet was born. I went up there to see his work, and once I saw the waveforms, I was convinced. There was much faxing of waveform printouts for a few months, and some vendors were rather unhappy, but the problem got fixed.
So that's why.