Friday, October 30, 2009 Helping Create the Next Token Ring?


That’s right. I said it—“token ring”. Eight oh two dot five. Everyone’s favorite token passing architecture, unless you were a MAP/TOP guy; if you were, I’m sorry.

Why am I thinking about token ring? Two reasons.

First, I’ve just spent three days at the P2030 (IEEE’s smart grid interoperability workgroup) meeting, hosted by IBM at their Thomas J. Watson Research Center north of New York City. IBM was synonymous with token ring; in the forty year old Watson facility (which, aside from being beautiful, has aged magnificently well), I’d bet money that Types 1, 2, and/or 9 cables are still providing tensile strength, if not actually passing any data. Heck, even today, when I see certain numbers (8228, 3174), I can’t help but think of the elegant (and extremely socialist) nature of token ring. Ethernet allows for serious bogart-age of the contention-based transmission medium; token ring’s deterministic approach ensured that everyone would need to share somewhat reasonably, even if some stations took a few more hits on the token than others.

Second, I just read an article over at Light Reading on the forthcoming International Telecommunication Union Recommendation. I’d encourage you to read the article, which is very well-written but to me is way off the mark. The gist is that certain entities using technologies with which will eventually compete aren’t happy about a new specification coming down the pike.

This is newsworthy how?

Maybe I’ve been in this game too long. Maybe we should just cease development and innovation so that existing technologies can enjoy their hegemony without threat from advancements which, well, you know, make things better.

Maybe not.

Sure, entrenched players want to protect existing markets and customers. If I were the quoted “executive for a major U.S. company using MoCA” who “said, on condition of anonymity that his company is happy with what it has now and has no plans to deploy”, I’d be shouting from the mountaintops, too, in hopes that a new technology didn’t come along and relegate my existing solution to the back seat. Mr. Anonymous has a solution that works well. Coolio. He probably has tens or hundreds of thousands of set-top boxes, maybe even millions, with embedded MoCA chipsets. He wants to ensure the continued success of MoCA, in hopes that the technology won’t go by the wayside, as well as to ensure that any new technology (e.g., doesn’t whack MoCA. Right. I get it.

But, these protests are disingenuous, as are the protests mentioned by the HomePlug Alliance. So there are “tons of HomePlug AV products” in retail shops. Okay. Well, there were millions of ports of token ring in Enterprise America two decades ago, too. From 1991-1993, I had an office at Sears Tower, a building with which I had a love-hate relationship whenever I went from the 19th floor (where I was stationed) up to the 72nd for meetings, a trek which required multiple elevator lobby transitions. But, that isn’t what remains etched most deeply in my mind. As a networking guy, one of my most vivid memories is of the wiring closets on the lower floors, those which had the biggest and broadest footprint. Most buildings have a main distribution frame (MDF) on the main floor or in the basement, and a single intermediate distribution frame (IDF) on each floor.

Not The Tower. No, no, no. The Tower had four IDFs on each floor. If that ain’t bad-ass, I don’t know what is. As you’d expect, each IDF had vertical riser connections back to the MDF; horizontally, each IDF handled one-quarter of each floor’s distribution needs. Literally and figuratively, the common thread was IBM Type 1 cable, the big, thick, shielded black cables with the hermaphroditic connector. Seriously…look it up. I used to comment only half-jokingly that if the Type 1 cable was ever taken out of The Tower, it’d fall down, since the wiring seemed to be such a vital component of the building’s infrastructure—and it really was. Iron might’ve been the spinal column, but the Type 1 cable was the spine itself, the neurological pathway enabling The Tower’s heart to beat.

When I took up residence at The Tower in 1991, a newfangled technology called 10Base-T was just becoming popular; those of us at Cabletron were locked in a duel with our hated enemy Synoptics to land-grab as much of the twisted-pair networking market as possible. The development of twisted-pair transmission freed Ethernet (or eight oh two dot three, if you prefer) from the handcuffs of frozen garden hose, vampire taps, and 50 ohm terminators, so we were all eager to earn more than our fair share.

I recall walking (lumbering, actually) into customers and prospects carrying my MMAC-3 (now with improved FNB!), giving demos of this new twisted pair technology. Many customers were sold immediately. Others, particularly Big Blue shops, didn’t come along quite so easily. 18 years on, I can still remember some of the questions…
  • “How do you ensure that one station can’t take over all the bandwidth?”
    “Well, uh, we can’t.” (This was in the days before per-port switching, so CSMA/CD wasn’t just a good idea, it was the law.)

  • “So, if you can’t prevent one station from taking over all the bandwidth, how do you ensure that each station gets a reasonable share?”
    “Well, uh, that’s kind of self-policing in the CSMA/CD protocol…random back-offs give each user a chance to jump into the conversation, like if you’re on a telephone conference call.”

  • “So, if I’m not guaranteed a time-slice like I am on token ring, how can you guarantee response time for my latency-sensitive SNA applications?”
    “Your 3270 terminal emulation package will provide whatever spoofing and keep-alives necessary.” (Fingers crossed)

  • “Seriously?”
    “Wow, will you look at the time…lunch?”
My point here is that token ring was what it was all about, but time marched on. While token ring evolved from a 4 mb/s to a 16 mb/s specification, and while the dual counter-rotating 100 mb/s FDDI came out for those who really, really liked their data to run in circles, token ring didn’t win. Ethernet/10Base-T ended up being cheaper, easier to deploy, and had much broader appeal. Was it the best technology? That’s debatable—once the world moved from shared to switched bandwidth, the argument became kind of irrelevant. But, Ethernet had the broadest industry support, and its simpler architecture enabled the economies of scale so vital to any semiconductor solutions success.

This isn’t a technology argument, folks. This is a business argument. Yeah, Betamax and 1394 might’ve been good, even technically superior, solutions in their day—but they were each surpassed by solutions that were as amenable from a business standpoint as they were from a technical one.

Plus, as CopperGate notes in the article, none of these solutions are going away tomorrow. Hell, even token ring needed a decade or so to drift into irrelevance. MoCA and HomePlug (and other solutions based on the coax, phone line, and power line media supports) aren’t being thrown into the rubbish bin just because the ITU has defined a new technical standard. Many wars remain to be fought, most importantly on the business side. If is truly better than other solutions from a technical standpoint, it has a solid shot at success—but so did token ring, Betamax, and 1394. But’s truly worldwide appeal could (and should) enable semiconductor manufacturers to deliver compelling solutions in the volumes required to make ubiquitous across the planet. If and when that happens, THAT’S when incumbents really need to worry.

One final thought. HomePlug’s president states that he’s “disappointed that they think it’s a good idea to create a standard that is incompatible with what’s out in the market.” Hogwash. If that were the case, we’d still be using 802.4 on the manufacturing floor rather than industrial Ethernet. That’s right—eight oh two dot four, the token-passing bus, also referred to by some as “the worst of both worlds”. Sometimes you’ve gotta eat your young. The industry might, and I stress might, be doing so with Hurdles remain; end user products won’t be out for at least another year, likely longer. But the race is underway; as consumers, we’ll all benefit from products which perform better, at price points which should be lower than we’re used to. If the ITU hadn’t undertaken its effort to unify under a single chipset the three wires found in the home, we’d likely see years more of geographical isolation per technology, hindering existing solutions from ever achieving the critical mass so desirable in the semiconductor world.

This is a volume ballgame. Crank it up.

No comments:

Post a Comment