The requirements for the nationwide delivery of multimedia services on the proposed NII had not yet been broadly considered when the NMIS Project began in 1993. The aggregate traffic demands of delivering networked multimedia services to a broader user community were clearly going to require a different type of high bandwidth network infrastructure. Previous research had suggested technical alternatives in the architecture of networking infrastructures capable of supporting large scale multimedia distribution, but because of an absence of usage data and other types of concrete statistical information, this research provided an insufficient basis upon which to architect large scale multimedia servers for nationwide services delivery.

Accomplishments

The NMIS project supported the evaluation of potential scaleable distribution architectures to accommodate multimedia on the National Information Infrastructure (NII). The key technical objective of this effort was to understand the relative efficiencies of various network architectures for distribution of networked multimedia.

John Yanosy of Southern New England Telecommunications (SNET) and Brett Leida of the Center for Advanced Educational Services (MIT) identified considerations for the design and implementation of networked multimedia delivery infrastructures including Video Dialtone, Video-On-Demand (VOD), and applied networking (ATM and SONET). These activities produced an integrated, engineering-economic model for VOD and Video Dialtone systems. [35]

The model quantified costs for various VOD network architectures, such as a network with a centralized-video-server architecture versus one with a distributed video servers where content was split between a central video server and distributed cache video servers. The primary objective was to determine the relative cost efficiencies between these architectures by weighing server cost against network cost. Additionally, sensitivity analysis was conducted for critical variables, which included network and server capacity, service subscription levels, quality-of-service objectives, and industry cost trends. Yanosy and Leida concluded that an evolutionary approach would implement centralized servers during the early technology phase with a migration to the cache architecture during the latter years as demand increases and the video server technology drives down the learning cost curve.

Future Directions

The requirements of mass delivery of multimedia over the Internet are still not well established. Now that solutions to technical production and delivery challenges are emerging, the network research community must soon actively address the problems of scale and national deployment. These issues can be addressed with further infrastructure evaluation, similar to that described above but adapted specifically to the Internet and its protocols.