I don't think Gilder would be willing to include a router vendor in his report for the time being, if ever, given his belief in wavelength-switched long-haul networks being used in the future. Then again, AMCC, which makes a ton of chips needed to convert optical data signals into electrical ones that can be processed by routers, is on the list, so you never know.
Still, if Gilder were to go for a router vendor, my guess is that he'd chose a startup known as Hyperchip. George has often criticized network architectures that try to go with limited bandwidth and try to make it up by means of QoS provisioning (i.e. his choice of Mirror Image over Akamai). Hyperchip has taken a gung-ho approach to IP routing. They're working on some sort of low-cost, superscalar architecture that offers orders of magnitude of more processing power, all the while paying little attention to QoS software.
I'm still skeptical of the viability of such an approach, as both DWDM channel counts and wavelength bit-rates appear to be growing too fast for routing speeds to keep up, creating the need for QoS-based prioritization, even with the processing power gains Hyperchip is claiming to offer. Still, the company definitely seems like the type that Gilder would be interested in, assuming that he'd want to recommen a router vendor.
Eric |
