As vendors of so-called "god boxes" struggle for survival, a startup has emerged that wants to supply them with silicon.

Onex Communications Inc. came out of stealth mode today to tell the world it's developing chips that process Sonet, ATM, and Ethernet, changing from one protocol to another simply at the click of a mouse button (see Onex Unveils Integrated System).

The startup is suggesting that it can replace the guts of an IP router and an ATM switch and a Sonet digital crossconnect with a simplified system based on only two types of chips -- a "service processor" on the line card and a switch fabric on the switch card.

"We will build in native handling capabilities on each port, so it can handle packets, cells, and circuits in their original format," says Paul DeBeasi, Onex's VP of marketing. "We're not talking about circuit emulation, as you would in ATM."

The end result would require significantly fewer chips than today's systems, which are built from service-specific, piecemeal components, says DeBeasi. He estimates that his company's silicon will slash cost, power, and space requirements by a factor of ten.

The big idea is to allow carriers to migrate from whichever protocol is popular today, to whichever is going to be popular tomorrow, without having to change their equipment.

On paper, it sounds like a good plan. But delivering on promises is going to be tough, if history is anything to go by. Onex's success will hinge on that of its customers, the systems vendors developing multifunctional boxes. So far, however, god-box vendors have been spectacularly unsuccessful (see God is Dead).

Why is this? The main reason is equipment that tries to do everything runs the risk of doing nothing particularly well. Add to that the fact that different functions, such as Sonet and IP, are likely to be managed by different groups within a company, and you have a recipe for disaster. These groups don't talk to each other and are not prepared to make compromises for the sake of functionality about which they know little and care less.

How well Onex can implement different functions remains to be seen, as the company is not ready to talk about the specific speeds and feeds of its products. Details won't be released until October.

Another potential problem area for the startup is the fact that its architecture is based around a "service processor," a device that's a close relative of network processors (on the IP side of the family). Like network processors, the service processor needs extensive programming in order to do its job.

Onex intends to supply firmware along with its chips, but more often than not, a systems vendor may want to change the code in order to include proprietary functions. That means systems integrators will need to employ software experts.

DeBeasi says that Onex is well aware of the difficulties of dealing with software -- its executive team has a strong background in software engineering. To speed the acceptance of its product, it is planning to offer a range of more advanced software products.

For example, Onex's engineers have devised a text-driven interface so that the customer's test team can write text-based diagnostics and verification tools. That means the test engineers don't need to be C-programmers (although they may need to be experts in Onex's proprietary text language).The startup is also touting the fact that it provides a full system simulator, so that customers don't have to wait for the hardware before they can debug their software. It has also designed an OPNET-compatible traffic-engineering model, that will allow systems vendors to find the bottlenecks in their designs.

There are a number of other reasons for taking the startup seriously.

For a start, it has $30 million in funding from some serious backers, including Venrock Associates, TranSwitch Corp. (Nasdaq: TXCC), Star Ventures, St. Paul Venture Capital, and Signal Lake Ventures.

It may not be a coincidence that Venrock is also an investor in Polaris Networks, a startup developing its own god box. Polaris's big claim to fame is that its box is based on switching fabric that can handle different types of traffic in their native format (see Polaris Builds a God Box).

Apart from having serious investors, Onex has also managed to pull in some impressive partners that are contributing to the technology. Most notably, it's joined IBM Corp.'s (NYSE: IBM) "Rising Star" program, which means that IBM will do all its fabrication.

Onex has also snared some heavyweight partners on the software side of things. These are Wind River Systems Inc. (Nasdaq: WIND), a maker of a real-time operating system, and Tensilica Inc., which provides RISC-based libraries for Onex's chip. "It's the enabling technology that allows us to have programmability," says DeBeasi.

Onex was founded in May 1999 as a spin-off from TranSwitch.

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