Adaptive processing

A better way to pool computing resources with peer-to-peer computing

A better way to pool computing resources with peer-to-peer computing

DataSynapse, a US-based firm targeting the financial services market, takes what it sees as a less intrusive approach to P2P processing. Chief executive Peter Lee explains that his financial customers have huge amounts of computing resources that have been going to waste and they are willing to buy his software to utilise them more effectively, providing it does not affect their day-to-day operations.

"Traditional distributed computing efforts have been failures in part because the utilisation of the remote resource has been ill-behaved," says Lee, citing crashing systems, degradations in performance and intrusive background applications. "So when you come back to your computer, the software exits but it takes ages to reload your application. In a capital market dealing room, you cannot bother traders like that."

Lee has built adaptive processing techniques into his system and has foregone what he calls the "hub and spoke" approach, where a central server keeps close tabs on client-side processes. Instead, his system pre-emptively reschedules tasks, building redundancy into the network so that the same portion of a task is executed on more than one client. If an end-user resumes using a computer before a task is finished, that task is simply thrown away and continued on another client. The company also built the system to handle random additions to and deletions from the client base. He claims that the client is a lot smaller too - less than 1Mbyte in size, with no hard disc presence.

Being able to utilise the distributed computing power that has previously gone to waste could reap huge rewards for companies that decide to take the technology on, but it is still relatively immature and if mishandled could have huge ramifications for desktop computing infrastructures.

So enter into it carefully and ensure that you do not gain all those precious clock cycles at the expense of existing application performance.

Back to Share and Share alike

By Danny Bradbury

Read more on Business applications