Azanda Flips Its Chips

When it launched a little over a year ago, Azanda Network Devices was developing packet processing silicon for OC192 (10 Gbit/s) data rates and higher. Now the company has revised its plans significantly and is tooling up to tackle the OC48 (2.5 Gbit/s) market.

Azanda, which announced $19 million in funding this week, has decided to focus on standalone traffic manager chips -- ones that prioritize different types of traffic by putting packets into different buffers (see Azanda Collects $19M).

And it's out to prove that it isn't another Acorn Networks, a startup developing traffic manager chips that went under last year (see Acorn Launches Processor; it went on to die quietly).

To this end, Azanda has come up with what it hopes is a low risk product strategy, according to CEO Steve Dines. In his view, Acorn didn't go out of business because it had the wrong product, but because it didn't position the product correctly. "It's not just what you've got," he says. "It's how you deploy it."

The packet processing scene has changed a great deal in the last few years, says Dines. Systems vendors started out thinking that their existing OC12 (622 Mbit/s) network processors would be able to handle OC48 with only a few minor adjustments. They didn't anticipate needing a lot more processing muscle until OC192 data rates were reached. At that point, they planned to provide OC192 boosters in the form of co-processors, such as traffic managers, that could offload specific functions from the main network processor. That was around the time Azanda was founded.

It turned out, however, that even the new network processors designed around OC48 couldn't process data at full line rate. As a result, deployment of OC192 was pushed out. Neither startups nor established chip makers could afford to wait for it, so they started developing co-processors that worked at the lower line rate. This explains Azanda's shift in product focus.

However, this doesn't go far enough to explain why Azanda might succeed where Acorn failed.

Dines says there are two key aspects to Azanda's "risk-free" strategy. First, it has designed a product that will appeal to existing players, rather than to startups. Put simply, it offers a chip that will work with any network processor and any switch fabric already out there. That way, a carrier can use Azanda's product to upgrade a single line card, instead of needing to install a complete new system.

"It is typical for a startup to design a product that only appeals to other startups," he says. "But that results in risk squared." Even working with an established company doing a ground-up design is risky, because the time to market is so long. Azanda's approach is safer because "you can predict the time to market for an evolutionary line card."

Indeed, most other packet processing startups and even some established chip makers have forced themselves into a corner, Dines contends, by developing traffic manager chips that can only work in conjunction with their own network processors and switch fabrics. Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC) is one example -- it uses a proprietary interface between its network processor, traffic manager, and switch fabric.

Although Acorn, like Azanda, was positioning its traffic manager as a standalone device -- one that would integrate easily with chips from other vendors -- its design wins were with the sorts of customers that take longer to get to market, Dines suggests, boasting that Azanda's six design wins are all with tier-ones.

The other big difference from Acorn, says Dines, is that Azanda has investors with deep pockets, as shown by the recent funding announcement. They include: Bessemer Venture Partners, Commonwealth Capital Ventures, Goldman Sachs & Co., and Highland Capital Partners. In his view, that's crucial, because many investors are working hard to keep their existing portfolios alive and have very little cash to put into new ventures.

So, dead companies aside, who represents the competition among the living? Probably Azanda's closest competitor is Vitesse Semiconductor Corp. (Nasdaq: VTSS), which introduced its second-generation traffic manager, called PaceMaker 2.5, last fall (see Vitesse Offers PaceMaker, TeraStream). "[Azanda is] looking at some of the same sockets that we are," agrees Jon Stilwell, Vitesse's VP of product marketing.

Azanda is claiming that its product will be better than Vitesse's in terms of power consumption, density, and something called "shaping accuracy," which determines how accurately the chip allocates bandwidth to traffic from different customers. "Azanda's stuff is so preliminary, it's hard to comment," says Stilwell in response. He points out that Vitesse has had chips in the market for over a year now, while Azanda says it does not expect to ship samples until May.

"If they can leapfrog us and do better, then they'll win some market share," he adds. Of course, Vitesse isn't standing still either -- Stilwell says it plans to reveal a new traffic manager chip next month.

— Pauline Rigby, Senior Editor, Light Reading
http://www.lightreading.com