Server: Netscape-Commerce/1.12 Date: Tuesday, 26-Nov-96 00:07:17 GMT Last-modified: Friday, 18-Aug-95 18:01:08 GMT Content-length: 3262 Content-type: text/html
![]() |
![]() |
![]() |
David D. Clark, Senior Research Scientist | John T. Wroclawski, Research Scientist |
Karen R. Sollins, Research Scientist |
The Advanced Network Architecture group defines network protocols and methods for the information infrastructure of the future. Guided by "real-world" requirements, we intend to influence all levels of network design through engineering studies, prototype development, and observation of existing networks.
The group's Information Mesh project, for example, is developing a network infrastructure that finds and relates information, rather than simply moving it around. In this view, the network-application interface accesses a global "mesh" of data and relationships that may be reviewed, searched, and manipulated. Individual data elements are controlled through specified access and use policies. The infrastructure presents a uniform interface through mechanisms that locate, retrieve, and smoothly integrate these data.
Naming and linking architectures are critical to the mesh design. The former, based on a model similar to that of human society, uses hints and locality to handle trillions of objects whose lifetimes range from seconds to centuries. The linking architecture, meanwhile, captures the web of relationships that grows as information is used.
Tomorrow's networks also will demand better management of bandwidth and buffering resources. This is crucial to the development of integrated service networks, which combine telephony, data, and entertainment. We have proposed a new network service model (with appropriate algorithms) that offers high-quality service for a wide range of applications. The Integrated Services Internet project, a collaboration of several research organizations, is currently developing standards and techniques to deploy this model on the Internet.
Network protocol designers also must provide higher performance, security, and flexibility. Our Application Level Framing protocol framework meets these demands by separating the protocol processing unit (used for error recovery, flow control, and similar management functions) from the multiplexing unit (used to place data on the communication channel). The separation leads to a simple, modular style of protocol design in which each application uses only the required functions. Results include improved performance and a better match with the demands of evolving applications.
The group collaborates with academic and industrial partners and participates in such efforts as ARPA's DARTnet project and the Aurora gigabit testbed.