Anjana Susarla

The Internet’s phenomenal impact and subsequent growth stemmed primarily from the open, public, worldwide networking it offers. Most current applications, however, use client-server technologies that incorporate networking as an ancillary, value-added feature. Peer-to-peer computing offers a radically new way of isolating and focusing on the networking aspect as the business model’s mainstay. A P2P network distributes information among the member nodes instead of concentrating it at a single server. This paradigm offers exciting advantages in information sharing, but it also presents challenging disadvantages. P2P networks differ markedly from the client-server interactions that typify applications in the TCP/IP world. A client-server scenario like the Web depends on a single server storing information and distributing it to clients in response to their requests. The information repository remains essentially static, centralized at the server, and subject only to updates by the provider. Users assume a passive role in that they receive, but do not contribute, information. A P2P network, on the other hand, considers all nodes equal in their capacity for sharing information with other network members. Each user makes an information repository available for distribution, which, combined with anyone’s ability to join the network, leads to the fast growth of a network composed of distributed information repositories. In a P2P model, each member node can make information available for distribution and can establish direct connections with any other member node to download information. Instead of looking at what is available in a centralized repository, such as MP3.com, a client seeking information from a P2P network searches across scattered collections stored at numerous member nodes, all of which appear to be a single repository with a single index, such as Napster. Member nodes primarily use existing P2P networks, such as Gnutella or Napster, to share information of some kind. The TCP/IP Internet typically admits these nodes without restraint, and any node that can access the network can join. Networks can refine admission policies, however, to restrict membership to interest groups. Likewise, the operation domain can be a controlled environment such as an intranet rather than the public Internet. P2P networks can use protocols that make it easier for individual nodes to participate and share information, but the trade-off can be decreased security or quality of service. For example, using a lightweight alternative such as the user datagram protocol may sacrifice the reliability that a protocol such as TCP offers. Manoj Parameswaran University of Maryland