Once upon a time there was a computer. Then there were many. Eventually it seemed like a good idea to hook them up. And so the Internet was born. So far we’ve considered the growth of the Internet in close relationship with the number of computers connected to it. This may all be about to change. On Feburary, 2008 IBM announced a new research endeavor: “to explore the construction and implications of a global-scale shared computer capable of hosting the entire Internet as an application.” According to the white paper linked above this monster will, at hardware limit, house 67.1 million cores running on 32 Petabytes (~ 32,000 Gigabytes) of memory. IBM hopes this undertaking will change the way the Internet currently exists, from poorly planned clusters of commodity computers to a more efficient SMP (Symmetric MultiProcessing) design.
I will, however, steer away from the technical complications that Kittyhawk must overcome to achieve it’s ultimate goal, and analyze a world where Kittyhawk does exist, where the entirety of the Internet is being run off a single computer. Some thoughts and observations. IBM headquarters would require security rivaling (maybe exceeding) nuclear silos. Moreover, a centralized implementation of the Internet puts redundancy measures into question: what if Kittyhawk fails? Granted it will probably have an innumerable number of fail safes, but reality still dictates for some chance of failure. IBM would probably, eventually stand trial for monopoly (particularly from the European Union). But ultimately we have lost the unending scalability that so characterized the Internet in the first place. Internet growth would now happen in steps, as Kittyhawk2 and it’s successors make it out of the lab. There is undoubtedly much to gain from this research for the sake knowledge itself. But if the decision to implement this replacement is ever on the table, would it be worth it?
* You can follow any responses to this entry through the RSS 2.0 feed.