Is Networking a Commodity?
In my last post, I left you with a simple point: When choosing a solution, the right solution is the one that supports the business. Find the business problem, scope the business problem, solve the business problem. When done, repeat.
It seems, based on this, that all businesses care about, in terms of the network, is the ability to move packets. To use a comparison that is often made: It might be “nice” to drive a “nicer car,” but in the end, a car is a car is a car. All that matters in cars is that they get you from point A to point B, wherever those points might be. If all the data a business cares about can be packed up into packets, and all that matters is getting them from point A to point B—wherever those two places might be, then the kind of equipment you use to move packets does not matter. Just like with cars, what you need is something to get the job done—everything beyond that is a luxury, and you are paying luxury prices for stuff you do not need. In other words, it seems that if the main driver of network engineering is solving business problems, then the network must be a commodity.
From a hardware perspective, networks certainly seem to be moving in the direction of being a commodity. Much like the server market, there are a few mainline chip makers making chips you can purchase wrapped in sheet metal painted different colors, with few differentiating factors. Software might seem, at first, to be the real differentiator, but open source solutions are quickly becoming viable for solving some 80% of the problems networks really need to solve. The other 20%? Well, a clever designer can work around those by throwing bandwidth at the problem, through automation, or… Software Defined Networks (SDNs).
If networks are truly becoming a commodity, the impacts will be huge. Network engineers will need to retrain into other fields, or just accept commodity paychecks. Network operators will need to need to find another way to make money, or accept the lowered profit margins of a commodity business model.
Before accepting this argument, however, it is worthwhile to look a little more deeply. Returning to the example of a car, are there really only two kinds of cars—vehicles that get you from point A to point B, and luxury vehicles that get you there in luxury, and you pay a luxury price for? Are pickup trucks, minivans, sedans, off road vehicles, and large over-the-road haulers (“18-wheelers,” or “lorries,” or some other description) really all the same? Extending the metaphor, is a plane the same as a car, or a train, or a ship?
No, clearly this is not true. Even if the four-door sedan could solve 80% of the problem, it will take a bit more than some clever design to replace a train with even a large number of four-door sedans. The argument, then, at least needs to make room for the variety of vehicles, and allow for those networks that will not, ever, fit into the commodity mold. These are not mere luxuries, these are designed to solve specific purposes.
But it is useful to move beyond even this, and consider the larger problem of transportation. Delivering a person from point A to point B does not just involve a car. It also involves a road system, including traffic signals, queueing design, scaling factors, and parking lots. The infrastructure that makes the car possible may consist of many different commodity components, but it is not, itself, a commodity. It is (or at least should be) a well-designed system of systems.
Is a network truly like a car? A commodity that is easily replaced? In the best designed transportation system, the answer is yes. Is the network itself, the system that moves not just packets, but streams of data across a long distance, a commodity? Clearly the answer is no.
The point of a network is to carry information efficiently from point A to point B—this much is correct. But the variety of information, the reliability with which it must be carried, and the speed at which it must be carried, provides something much more than a mere commodity to the business with vision.
I remember once being told by a business leader that the network is a commodity, and the business was foolish to invest in it. Then I asked a series of questions. How many customers do you have, and where? What do those customers want? How many of each item should you make, and why? Ultimately, all businesses are in the business of information. And networks move information.
What lessons can the network engineer—and the business leader—take from this?
First, focus on the network as a system, rather than the network as individual devices. The closer the individual components of a network come to being commodities, the more the system should be the focus, rather than the device. This is the shift network engineering is undergoing right now. Individual network engineers can either move towards the system, or towards the commodity.
Second, focus on what the business needs, at a systemic level, and how to solve those problems. Yes, find the business problem, scope the business problem, solve the business problem, repeat. The network is only a commodity when it can add no value to that process—but so long as the business relies on information, there will always be some way in which the network can add value.