Computing That Serves

The Hyperweb

Dr. Scott Woodfield

Students working with Dr. Scott Woodfield have created the Hyperweb, an information sharing infrastructure that reduces the problems seen in the client/server model. Their goal is to increase reliability by a thousand fold, decrease provider costs by two orders of magnitude, and have the ability to scale to billions of users, while being as efficient as any client/server model.

In the client/server model, the architecture is unreliable because the server presents a single point of failure. When any one of a variety of conditions, including power problems, communication failures, hardware faults, and overload conditions, cause the server to rail, all clients become disconnected.

The Hyperweb, however, is a peer-to-peer system created from a structured interconnection of nodes. In the Hyperweb, each client, or node, is connected to log2(N) nodes, where N is the number of nodes in the Hyperweb. This structure makes the Hyperweb extremely reliable-if one group of nodes fails, millions of other nodes will still support the Hyperweb. Furthermore, because there is no centralized server (each user gives a small amount of resources to a decentralized server), the information provider has no extra overhead. The clients incur no additional expenses either, except for the use of a little more disk space.

The Hyperweb compares very favorably to other peer-to-peer topologies. It has shown itself to be more reliable, faster, and more efficient than other systems.


Website address: