Ellen Zegura is Professor and Chair of Computer Science at the Georgia Institute of Technology. She writes to us today in her role as chair of the NetSE Council.
What is a “better Internet”? The current Internet has been a remarkable success, providing a platform for innovation that far exceeds its original vision as a research instrument. It is well documented that the Internet has transformed the lives of billions of people in areas as diverse as education, healthcare, entertainment and commerce. Yet many of these successes are threatened by the increasing sophistication of security attacks and the organizations that propagate them. A materially more secure Internet would be “better”. Further, billions of people remain untouched by the advantages of the Internet; Internet World Statistics puts worldwide average Internet penetration at about 22% in mid 2008. An Internet that affordably reaches the other 80% of the world population would be “better”.
Beyond security and accessibility, there are other areas where limitations of the current Internet are significant. The Internet usually works pretty well, but every user has experienced inexplicable periods of degraded performance or outright non-function. The current Internet provides no visibility to end-users and shockingly little visibility to network managers and operators to support understanding, adapting to and fixing reliability problems. Such limitations require lay people spend their leisure time as network systems administrators and companies to spend heavily in network operations. Further, the lack of performance reliability prevents the Internet from advancing to become a truly dependable, critical infrastructure. Indeed, current societal reliance on the Internet for critical functions is disproportionate to our ability to deliver a high degree of dependability. A more predictable Internet would be “better”.
The Internet embeds societal values in ways that are often implicit and not well understood. For example, the Internet is “open”, usually intended to mean that anyone can join the network by implementing the public protocol IP. In principle, users can run any application on the Internet, without limitation imposed by the network protocols. Open networks promote organic growth, but suffer from a lack of mechanisms to vet or bar participation. Issues of trust and individual accountability are confusing. As the well-known cartoon says, “On the Internet, no one knows you’re a dog.” An Internet that contains support for identity would be “better”.
The research community is poised to dramatically advance the agenda of building better networks through advances in both empirical design methodology and systematic design methodology. We have an approach to support large-scale and flexible experimentation based on programmability of devices and federation of multiple test-beds. We have a nascent mathematical framework for understanding architectural features and underlying principles. The time is right to advance and link both methodologies to realize better networks.
I find this article interesting, especially because it raises issues beyond security, such as management and robustness issues. New York times did an article focusing mainly on security. It elicited mostly negative responses on Slashdot, primarily because it seemed to suggest that you have sacrifice privacy or worse to achieve security. I wonder what we can do as a community to explain the trade-offs better, assuming we know what they are.
I agree that there are interesting issues beyond security and that tradeoffs are really the key interesting question. Security faces other tradeoffs besides privacy, including openness and expense.
The Internet and the devices that connect to it are consuming an increasingly large amount of electricity. It is estimated that in the US alone this electricity consumption is several percent (of the total electricity consumed) at a cost of tens of billions of dollars per year. Much, if not most, of this consumed electricity is wasted due to the non-proportional energy-utilization behavior of network links, equipment, and end hosts. There is tremendous opportunity for significant energy savings *in* the Internet. Such savings could be achieved by energy-aware protocols that would enable network links, equipment, and end hosts to sleep more. There are even greater opportunities for energy savings *by* the Internet enabling smart building, smart grids, and smart “stuff” in general. A Green Internet that is energy efficient itself and also suitable for enabling energy efficiency of modern society would be “better”.