Why Azul is bullish about network-attached processing systems

By John Gallant, Network World
January 07, 2005 09:43 AM ET

. What's this?

I feel sorry for the person who can't get genuinely excited about his work. Not only will he never be satisfied, but he will never achieve anything worthwhile.
Walter Chrysler

Dear Vorticians,

A happy and prosperous new year to you!

You know, I'm a sucker for enthusiasm. Over nearly 22 years covering the technology industry I've had the pleasure of meeting many entrepreneurs and savvy IT executives. I've always been captured by the energy and zeal of the real innovators - the people who want to set the industry on its ear.

Related Content

Over the past few years, it's been tougher to find such groundbreakers. With the enterprise IT field in rehab, there was far more emphasis on cost-reduction and incremental improvement vs. real, bet-the-barn kind of innovation.

But, in the past couple of months, I've had the pleasure of conversing with two CEOs who are setting out to do nothing less than change the tech world. Their goal is to not only capitalize on the new developments in enterprise IT - changes that we've chronicled in this newsletter and the VORTEX conference - but to knock the current leaders, the big boys, right off their perches.

This week, I want to introduce you to Stephen DeWitt, CEO of Azul Systems. When I spoke with DeWitt this week about Azul's so-called network-attached processing systems - more on what that means in a second - he started off by saying that Azul's launch will be "one of the biggest stories of 2005." Later, he told me that what Azul will deliver will "shatter the economics of computing." No wallflower he.

Now here's what lies behind this bravado. DeWitt says that traditional computing architectures - like x86 or Sparc-based systems - are not designed to deal with new applications built with Java, J2EE or .Net (virtual machine-based applications). These applications are very demanding and they don't behave well enough for users to get any real scalability out of today's machines.

Grids, virtualization and on-demand, according to DeWitt, are attempts to mask the limitations of those traditional computing architectures, but they are complicated and clumsy ways of solving the new application problem.

So, the folks at Azul Systems are branching off in a completely new direction. They've built a box called a network-attached processing system that utilizes completely new multi-core microprocessors. Each chip has 24 processors on it and each box can house up to 16 chips, for a total potential capacity of 384 processor cores, along with 256G-bytes of memory.

Linked to existing servers, the NAP enables the virtual machine based apps to call on a huge pool of computing power as needed, attaining what DeWitt claims is linear scalability. (By the way, DeWitt is the former CEO of Linux systems vendor Cobalt Networks, which he sold to Sun for a sunny $2 billion in the shiny days of 2000.)

It sounds too good to be true, but DeWitt says the NAP is in customer sites today and the beta tests are going swimmingly. We've heard that before and we remain professionally skeptical, though personally excited by the prospect that it could actually be true. I'm not sure why I slipped into the royal we there, but as my mother is fond of reminding us, there's many a slip 'twixt the cup and the lip.