What would you do if computing resources became free? That's essentially the question that Chris Anderson asks in his Long Tail blog. You see, MIPS are quickly becoming a free resource in the world of computing. Dual-core processors are now relatively common. The future promises CPUs with tens of cores on them in just a few years. Indeed, we may see a Moore's-law-like effect where the number of cores per CPU doubles every two years.
How does this affect Vyatta and open source networking? Well, it suggests that more heavy weight problems can be solved with standard hardware, and for less money than in days gone by. Imagine a standard CPU where five out of ten cores are performing packet forwarding while the other five cores are doing things like VPN encryption, anti-virus scanning, and IDS processing.
Some people will be quick to note that while MIPS have been made increasingly free, memory bandwidth is still a bottleneck. That's true. Caches have been made larger and larger to help compensate for the gap between MIPS and memory bandwidth. Unfortunately, for some problems cache doesn't do you much good. Further, networking is one of those problems. Remember that operating on packets, at least for basic forwarding, is a one-time operation; every packet is new (and therefore not in cache), and then it's gone, never to be seen again.
Fortunately, there are things that are happening to help increase memory bandwidth (DDR2, RDRAM, and faster front-side busses) as well as reduce latency (integrated memory controllers ala AMD Opteron/Athlon). These innovations, coupled with general increases in I/O bandwidth (PCI Express) will help standard hardware do more and more.
Send me your thoughts at "dave at vyatta".