What a tera-ble idea!

It's begun. Just when you're getting used to the idea of gigabit Ethernet, along comes the idea of Terabit Ethernet. There is talk of industry consortiums. Venture funds are evaluating their investment options. And users are asking if they should bypass Gigabit Ethernet and wait for Terabit Ethernet later this year.

HotBot search on "Terabit Ethernet"

What have we done to deserve such cruel treatment? Isn't all the hoopla and angst surrounding Gigabit Ethernet enough? Still, given the number of questions I'm receiving from both venture-capital types and real live users, I thought it worthwhile to answer some of the questions surrounding this this new, exciting and non-existent technology.

Q. Does Terabit Ethernet exist?

No. This is a bizarre form of mass hallucination that should be avoided (like most other forms of mass hallucination).

Q. But aren’t some vendors speculating about Terabit Ethernet as the next upgrade to Gigabit Ethernet?

Yes, some vendors have decided to bring that subject into the open. And yes, they should be banned from public for doing so.

Q. But what about all the talk about "terabit" products?

Avici offers glimpse into its really big router Terabit Switch Router. Network World, 1/12/98.

While Terabit Ethernet does not exist today, we have seen a wave of product discussions around switches and routers that can perform at the terabit level. But this is very different from "Terabit Ethernet" - in most cases, the underlying technology used for I/O is either Gigabit Ethernet or ATM.

What about the requirement for Terabit Ethernet? Aren’t network speeds continuing to increase?

The question with Terabit Ethernet is not one of "can it be done" (it probably will be some day), but rather "is it required anytime in the next year or so?" The answer to that is probably "yes" only if you're planning company-wide Quake marathons or if you plan on downloading the entire Library of Congress this afternoon.

Gigabit, shmigabit Online forum on whether most shops even need gigabit connections to their servers.

Otherwise, the demand for Terabit Ethernet (a single link communicating at terabit speeds) is not yet here – we are having a hard time filling up gigabit speed links, not to mention the difficulty we now see in deploying Gigabit Ethernet products. Further, while the leap from Fast Ethernet (100M bit/sec) to Gigabit Ethernet (1,000M bit/sec) was relatively simple, the leap to Terabit Ethernet (1,000,000M bit/sec) is more than just an issue of wire speed.

How many switches out there can support more than a few hundred Gigabit Ethernet pipes today? I can’t think of any right now. By mid-year there will be a few of the early switches that reach up into the multiple hundreds of Gigabit Ethernet ports, but it won’t be until late 1999 that we really begin to see any devices that could actually utilize a Terabit inter-switch link.

Q. How would such a switch be built?

Exactly how is still under speculation. But it would clearly be very expensive. There is no commercial technology today (i.e., silicon) that can switch or route at that speed. To create terabit switch/routers, we will likely see multiple Gigabit Switch implementations bundled together, much like we see in the carrier switching world. However, even if we were building and purchasing Gigabit Ethernet (or ATM) switches with the aggregate capacity to switch more than a terabit of information, I’m not convinced that Terabit Ethernet would be the most efficient mechanism for the inter-switch connection. It seems to me that ATM or a Frame/IP over SONET is much more likely to be the commercially viable technology.

The interesting question is who would deploy such a beast first. Will it be the carrier/ISP organizations looking for the next big core switching platform? Already some of the carrier-class optical switches are in this range. Or will it be users hungry for a mechanism to bundle together all of their newly implemented Gigabit Ethernet switches?

My bet is on the carrier/ISP side of the market. Unlike the Internet, which shows little signs of a slowdown in the need for bandwidth (especially with all those voice and multimedia apps coming online), the campus LAN is finally starting to show signs of catching up with demand. Networking vendors are already seeing a slowdown in the deployment of high-performance campus networks (due partly to lessened bandwidth requirements, but also to such issues as the diversion of IT funds to fix Y2K problems and even general slowdowns caused by the Asian financial crisis). Users are also growing more reluctant to trade up to the most powerful PCs (I can’t find a single user who is telling me that their desktop office applications just won’t perform on anything less than a 600-Mhz machine).

Q. If not Terabit Ethernet, what will be the next "massive" technology?

My bet is on an innovative technology developed by Steve Taylor, president of Distributed Networking Associates. Steve speculated several years ago on the promise of a new technology called Tachyon Transmission Mode (TTM). A tachyon is a theoretical particle that can travel faster than the speed of light. The beauty of TTM is that it operates so fast, that you could expect information to be received at the destination before it gets transmitted from the source (talk about a buffering problem!).

Q. So what should we do about Terabit Ethernet?

Nothing. Do absolutely nothing. When you see the word "terabit" in the press, just remember that what we need now is switches with overall terabit switching capacity, not another Ethernet standard.