Virtual Organizations, Pervasive Computing, and an Infrastructure for Networking at the Edge

That the Internet is undergoing explosive growth is not news. However, how that growth is happening might come as a surprise to many. We are quickly reaching the point (if we have not already) when the majority of nodes on the Internet are not computers running browsers or web servers that are directly used by humans. Instead, the Internet is becoming the communication fabric for devices to talk to services, which in turn talk to other services. Humans are quickly becoming a minority on the Internet, and the majority stakeholders are computational entities that are interacting with other computational entities without human intervention. Such change brings significant challenges and opportunities to those who wish to use or aid in the use of this connecting fabric. In a world of humans browsing the net, we could design our network-available services assuming that the use and applicability of those services could be recognized by something with human intelligence. When services must be recognized and used by other computational entities, no such assumption can be made. The intrinsically human abilities of understanding the meanings of descriptions and being able to figure out how to interact with a service are not currently within the scope of computational entities, and may never be. With a human in the loop, we have been able to go far by using the well-understood client-server model of distributed computing. But as the human moves out of the loop, we enter into the realm of peer-to-peer distribution, an area in which our understanding is far more limited. At the same time that we are switching to this new model of services, we are also expanding the set of client devices that are used to access services over the network. We can no longer assume a pc-class device running a full browser on a megapixel bitmap screen. Instead, the services on the network will be accessed by cell phones, palm devices, home appliances, and embedded systems in our automobiles. Some of these devices will be stationary, but many of them will be mobile. The set of devices will be constantly changing, and the protocols needed to talk to these devices will be constantly evolving. Our current models for system administration are insufficient for this environment, and new models will need to be adopted that allow dynamic evolution and rapid change. This paper will center on the new problems that will be encountered as we go through this change in the Internet, and discuss some of the more promising approaches to the problems. In particular, we will see how a combination of traditional techniques in distributed systems and some of the technologies traditionally associated with agents can lead to a network that is self-administering and allows the kinds of rapid change and evolution that will be required if the Internet is to continue to grow and thrive.