Azul jumping into compute pools

The concept of providing businesses with compute pools of network-attached processing power is the brainchild of Azul Systems, a Mountain View, Calif.-based start-up.

This week, Azul executives, including its president and CEO, Stephen DeWitt, former CEO at Cobalt Networks and a former vice president at Sun, are unveiling their first product and inviting corporate IT to jump in. As they explain it, a compute pool contains the CPU and memory part of the computer.

Although the terminology and technology is new, the idea is to 'solve pain points' in business computing that are at least as old as the error message advising 'insufficient memory,' says Azul co-founder Gil Tene, the firm's VP of Technology and CTO. It is also a variation on 'the network is the computer,' as it provides appliances that plug into networks and automatically supply processing power to network applications on an as-needed basis.

The Azul compute appliance solves the same resource problems as Grid and clustering, Tene tells eADT, but with a more 'simple and elegant solution.'

The company plans to offer network-attached processing along the lines of network-attached storage, he adds. Azul has developed a resource management system that allows seamless mounting of compute resources in a virtual machine environment.

To explain why he is targeting VM, Tene points to Gartner estimates that 80% of development by the end of 2008 will be based on VM environments, whether is it J2EE or .NET.

He estimates that in large corporations today, IT departments cope with an average of a new Java app coming online every 24 hours. When he asks potential customers how they are planning for the additional computing resources those apps require, he says the answer is usually: 'Badly.' Tene finds the best rule of thumb in IT comes down to 'add an app, add a server.'

If the Azul product works the way Tene says it does, it would allow IT to set up racks of compute appliances to form a pool of CPU and memory that would be automatically allocated to whatever app or Web service needed it.

Tene says Azul's appliance plugs into gigabit Ethernet and does not require any modification to the app. In addition, it does not have 'binary compatibility requirements or operating system dependencies,' he notes.

This kind of resource was not possible before Azul designed software, and especially chip-level hardware, capable of providing the needed computing power. The concept of mounting such power wasn't enough, he says.

'We had to build power worth mounting ... based on silicon we built ourselves,' he explains.

That power comes from what Azul describes as 'multi-core silicon technology' with each compute appliance containing '384 coherent processor cores and 256 gigabytes of memory in a fully symmetrical multiprocessing [SMP] system.'

The Azul product is set for field-testing later this year, and for release in 2005. More information is available at

About the Author

Rich Seeley is Web Editor for Campus Technology.


Sign up for our newsletter.

Terms and Privacy Policy consent

I agree to this site's Privacy Policy.

Upcoming Events