Startup Procket Networks Inc. is known for holding its cards tightly to its chest, but Light Reading has managed to gather some new details about its strategy and specifics about its products.

First, the product strategy. The company is not focused on building either an edge routing product nor a core router, company sources say. It now plans to introduce an entire suite of routing products, says Ron Young, chief marketing officer for Procket. This is a shift from the company’s original plan, which was believed to center on building a high-end core router to compete against Cisco Systems Inc. (Nasdaq: CSCO), Juniper Networks Inc. (Nasdaq: JNPR), Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7) and the half-dozen other startups going after this market.

“The company was originally aiming narrowly at one product category,” says Young, who wouldn’t reveal specifically which product categories the company is currently targeting. “Now there is a family of products. Why do it this way? If you put this much energy into developing a technology, you should be able to leverage it across multiple revenue streams.”

Former employees say the cornerstone of the product line is a 12-slot core router chassis that uses highly dense, scaleable chipsets custom-built by IBM Corp. (NYSE: IBM). This box will offer 960 Gbit/s of capacity (480 Gbit/s in and 480 Gbit/s out) in half a telecom rack.

If this is true, the new Procket core router represents a leap in density from currently available Cisco and Juniper gear. Juniper’s T640, currently the highest-capacity core router on the market, supports 640 Gbit/s of capacity in half a rack (see Juniper Goes Terabit With the T640). Cisco’s GSR 12416 only supports 320 Gbit/s of capacity in one full telco rack. Juniper plans to scale its box by adding an optical matrix. Cisco supposedly has a next-generation router due out later this year called the GSR 12816, which will scale to 1.28 Tbit/s (see Cisco's Got a Terabit Router Too).

As mentioned above, Procket doesn’t plan on using this technology only in the core. It will use the same boards and chips found in the core router to build smaller boxes with less throughput and fewer line card slots.

So where will these boxes be used? Procket's high capacity approach looks more like something built for the core, despite the speculation from some observers that the company is building routers for the enterprise (see Procket's Got a Two-Pronged Plan ). Others say it is targeting the edge router market (see Is Procket Heading Toward the Edge?). While Cisco has been successful in addressing both enterprise and service provider needs, this is not an easy task for a small startup. And as Juniper found when it tried to storm the edge-router market with its M10 and M5 products, it’s not always easy to scale down a core router to address the edge.

Juniper’s M10 and M5 edge routers were based on the same technology found in its flagship M40 router. But edge routers require a different set of software features. As a result, Juniper bought Unisphere Networks for its ERX edge routing product (see Juniper Nabs Unisphere for $740M).

“I don’t consider Procket’s smaller boxes edge routers,” says one former Procket engineer who was part of a recent company-wide layoff. “They won’t support all the fancy software features found in a true edge router.”

Where are these lower-capacity Procket routers likely to find a home? One place could be in smaller ISP networks that don’t need the capacity of the high-end product. These routers can also be used in the so-called "metro core" for aggregating traffic and feeding it onto the Internet backbone.

"They are definitely building a core box and all this talk about the edge is really about them filling out their product line," says Michael Howard, cofounder and principal analyst at Infonetics Research Inc. There are also other, less flattering, tidbits of news leaking out of the company. Just last week it had a company-wide layoff: Some reports suggest that as many as 40 people, or 15 percent of the workforce, was let go. Procket's Young admits that there was a layoff, but he says the number was much smaller than 40. He also insists that it was all part of a natural course of business, and he claims the company still has over half of the $272 million it raised from venture capital firms.

But things may not be going as smoothly as Procket officials imply. One ex-employee says that Procket’s products are behind schedule, with engineers working to fix two key problems associated with the hardware design of the box. One alleged problem is that the line cards are not hot-swappable. As a result, the company will likely offer a single blade version of its product first, says a former worker at the company. The other problem is that the boxes are not able to deliver line rate performance when all the quality of service (QOS) functions are turned on.

Young denies both of these claims.

“This is absolutely not the case,” he says. “Potential customers that are trialing the products are giving it rave reviews. We’ve actually had higher yields than we first would have guessed, and we are ahead of schedule.”

However, if these claims are true, Procket could have real problems.

“These are both serious design issues,” says David Passmore, an analyst with the Burton Group. “You have to architect a system properly from the beginning to deal both with hot-swappable components and to be able to handle processor-intensive tasks like QOS -- especially if you’re selling to the service provider market.”

— Marguerite Reardon, Senior Editor, Light Reading
http://www.lightreading.com