SAN FRANCISCO – You know you’re in trouble when Intel’s employees have a hard time explaining their own keynote material.
As Intel Developer Forum (IDF) 2002 wraps up Thursday, attendees will be walking away with many things – software
toolkits, business cards, hangovers. They’ll also be bringing back enough jargon to fill one of the many IDF complimentary backpacks that have become the conference’s most ubiquitous accessory. One of these buzzwords might actually turn out to represent a technological breakthrough, but the industry at large may never realize the full implications of it. It’s called Hyper Task Chaining.
Intel is introducing this technology in some of the network processors which were unveiled during the Tuesday morning session of the show. Many people have probably already forgotten about that keynote by now. They’ve since been dazzled by images of Bigwater, a concept PC Intel exec Louis Burns showed off that can be mounted onto the wall and offers expansion capabilities through a series of modules. They might also be recalling the image of an Intel employee showing a scene from “”The Lord of The Rings”” in 30 frames-per-second, full motion video on a handheld.
As for Hyper Task Chaining”” (I’ve capitalized the first letters because it is probably patented), the most we ever really saw was a crazy-looking foil in a slide presentation which showed lines criss-crossing between little boxes. Tom Franz, vice-president of Intel’s access and switching group, called upon one of his colleagues to explain the concept during his keynote. I could not follow a word of it, but for once this didn’t bother me, for I wasn’t alone. Even Franz jokingly looked back at us after it was over and said, “”Did you get that, audience?””
We didn’t get it.
As a journalist, I have the benefit of attending media-only briefings where Intel executives come to discuss new technologies in more detail. They try to dumb it down – way down – so that we can effectively explain it to the rest of the world. In the case of Hyper Task Chaining, however, our presenters were stumped. “”Hyper Task Chaining – Oh God. I can’t even BEGIN to explain Hyper Task Chaining to you in 15 minutes,”” said Intel exec Steve Price, who was dismayed at being given less than half an hour to talk about the network processor developments. “”Just take it from me that it makes it all go really, really fast.””
I’d like to do that, but then I wouldn’t feel like I’d done my job. So here goes: from what I’ve read and heard, Hyper Task Chaining somehow breaks down a packet or cell of data as it is moving through the processor. This allows the processor to handle more than one operation at the same time. The tasks are divided into “”function”” chains, in which multiple tasks are performed on the same packet, or “”context”” chains, where the same control structure is used on the same packet. It seems to bring a sort of order to the way tasks are executed as well – instead of the packets meandering all over the place to deal with things semi-randomly, the chains mean that one task is executed, then the next task. This is what speeds things up – to the point where the packet is out within about 35 nanoseconds.
Hyper Task Chaining has a second obstacle to broad industry understanding because it will probably be confused with hyperthreading, another key technology enhancement that has been much discussed at the show. Like its networking counterpart, hyperthreading is a sort of double-your-pleasure trick, making one chip operate like a dual processor. It is being introduced into the Xeon server chips and will be in the next-generation desktop chip, Prescott, as well.
Unlike some of the more PC-centric shows like PC Expo or Comdex, IDF demos are hard to dress up. In most of them, you’re basically watching two screens where a little metre or radar blip indicates how fast one chip is moving compared to the other. Hyper Task Chaining is probably a very useful and important contribution to Internet traffic management, but telecommunications professionals will probably only appreciate it once they get enough money to buy new equipment where the network processors use it.
A trip to IDF will show you that some innovations don’t belong on a stage, even if they do belong in a network. And sometimes, even when you’ve come all the way out here, you’ve still just got to be there.