IBM’s new take on processing

IBM’s initiative to open up the architecture of its Power RISC processors to third parties, announced this week, has interesting implications for networking.

At a press event in New York, IBM said third-party developers will now be able to customize Power chips for their own applications, and those applications could be networking-related.

Also, the Power architecture is going to be more flexible in other ways, which should make the processors faster and make their network aspect more dynamic. That’s because the architecture will now have the ability to change where it is putting its resources.

For instance, if a processor contains interfaces for Ethernet, USB and other I/O, those interfaces draw power and generate heat, even if they’re not really being used. IBM is making it possible for software to change the hardware – to make it so that the hardware in the chip is used for an application that needs it instead of just sitting there and consuming resources.

The example IBM gave me was that software could detect that only one of three Ethernet interfaces is really being used. Then, it could reroute that processing power for something else.

It ought to be interesting to see where this goes. IBM is trying to build a community of development around the Power processor – not in software development, but in Power chip development itself. Where will the industry take it – if anywhere?