10-Gigabit Convergence at Hand
On the face of it, Network Elements' "multiprotocol processor" appears to perform a similar function to the XENON chipset that PMC-Sierra Inc. (Nasdaq: PMCS) announced yesterday (see PMC-Sierra Touts Flexible 10-Gig Chip). These are the first chips capable of handling both the LAN and WAN versions of 10-gigabit Ethernet. They can also process OC192 (10 Gbit/s) Packet-over-Sonet (POS) signals, simply at the flick of a software switch.
These announcements mark the beginning of a new trend -- the convergence of telecom and datacom equipment. A driving force for this is simply the way the numbers add up: Traditionally Ethernet has advanced by a factor of ten with each generation, whereas Sonet has moved on by a factor of four. At 10 Gbit/s, telecom and datacom protocols have (almost) the same bit rate, for the first time.
As a result, equipment makers want one-size-fits-all line cards, to keep their box designs simple and flexible.
However, there are some crucial differences between developments at PMC and Network Elements. PMC-Sierra's chip is a product in its own right, while Network Elements has made a technology announcement. It won't sell the multiprotocol processor separately, but plans to incorporate it into a module that will be released later this year, in Q3.
When it becomes available, the module will be a complete replacement for a line card, plugging directly into the backplane of a box.
Bruce Murdoch, CEO and president of Network Elements says that doing things this way saves his customers a huge amount of time and money. "If you buy off-the-shelf parts, such as a laser and a laser driver chip, then chances are they don't work together [at first]," he says. "And even if they do, you still need to integrate them into a board, which takes time. Our customers are telling us that they save a year in time to market [by using modules]."
In terms of cost, the savings are harder to quantify, says Murdoch. But to give an idea of what's involved he refers to the physical layer (PHY) optical modules that Network Elements already sells. These cost around $5,000 each. It would cost an equipment vendor much more than that just to buy the bare parts that go into the module, and they would still need integrating onto a board, he claims.
Analysts feel this approach is pretty attractive. "I like the integration of optics with electronics because it allows the network to be more intelligent as well as more agile," says Jay Patel, a senior analyst with The Yankee Group.
The time-to market savings that a module affords could provide a runaway advantage. "Once a few customers start adopting modules, everyone else will have to follow, in order to keep up with them," says Murdoch.
Naturally, there are potential disadvantages to the module approach: The systems maker has to accept the module manufacturer's particular combination of components, and these may not be the best ones for the job in every case.
PMC-Sierra claims that its XENON product is more fully featured than Network Elements' processor. XENON supports more modes and interfaces, according to Steve Perna, VP and general manager of PMC's optical networking division. "This product announcement [from Network Elements] really is a very minor subset of all that XENON offers to builders of 10-gigabit metro equipment," he says. "Unlike XENON it is not an optimized chip-set architecture."
But that's not necessarily a killer blow to Network Elements. It could opt to buy multiprotocol chips from one of the traditional chip makers like PMC-Sierra or Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC). In fact, Murdoch says they have already been trying to sell him some chips.
This is also the reason that Network Elements chose Intel Corp. (Nasdaq: INTC) as an investor (see Nortel, Intel Back Subsystem Startup).
— Pauline Rigby, Senior Editor, Light Reading http://www.lightreading.com