Network Computingwww.networkcomputing.com

Many application developers have jumped onto the unlimited bandwidth bandwagon. I recently heard a consultant endorse the following as a planning exercise: Consider the ramifications to applications and infrastructure if bandwidth were unlimited. Clearly, this is meant as an exercise, but I don't ever recall hearing a consultant say, "How would you design if bandwidth were your most precious commodity?"

Application developers demonstrate their belief when they make application-architecture decisions that have drastic networking consequences, but never take those consequences into account.

Even IT networking professionals buy into the dream of unlimited bandwidth. A recent NetWorld+Interop dedicated a forum to the topic, and I've had many clients say they're thinking of simply provisioning their networks to have enough bandwidth to carry all their traffic. They imagine they'll never have congestion. Huh? These people probably haven't noticed that 10 percent to 30 percent (or more) of their IT budgets is going toward pumping up a carrier's income statement.

Nearly all vendors and networking service providers are dreamers, too. I recently had a discussion with a value-added carrier that based all of a client's application response time woes on the fact that the local loop wasn't fat enough and claimed that, in fact, the frame relay network could expand to unlimited bandwidth for that user. I think someone's missing the point here. It's one thing to build a network with unlimited bandwidth (an oxymoron), but try to sell one and you'll see that no client would ever purchase unlimited bandwidth. Not any single client, anyway.

Sometimes I think the only reason anyone wants unlimited bandwidth is that they're too lazy to calculate how much bandwidth is enough. Capacity planning is certainly more an art than a science these days because there are so many more applications with so much more unpredictability. But it's something IT must do.

If you ask me, unlimited bandwidth won't ever happen, at least not for the purchaser of bandwidth. You'll only have (and this is still a vast simplification) unlimited bandwidth on the WAN if you have unlimited money. How many folks do you know who fit that description?

Granted, prices are falling, and pricing strategies are gradually rendering distance meaningless. Frame relay PVC (Permanent Virtual Circuit) pricing often is not sensitive to distance. Similarly, the Internet can get you across the world at $20 a month--just not very fast. Unlimited time is not the same as unlimited bandwidth.

Bandwidth still costs money. Until we have free bandwidth, we can't begin to think about having unlimited bandwidth. Yet free and unlimited bandwidth remain elusive goals. We continue paying our monthly tithe to the carrier gods. The amount we paid last year might buy us more bandwidth today, but by now we've determined our bandwidth needs have expanded. Unfortunately, the more we can get, the more we want. What we have is never enough.

More of Everything Over Time Remember Moore's Law? It states that the number of transistors per square inch on integrated circuits doubles every 18 months. The corollary to this law is that the cost of MIPS is falling at a steady (and astonishing) pace. That doesn't mean that computers cost less (though this may be true), but it does mean that the same money, say $2,000, buys a much faster PC every year or two.