Buffer design for wide-area ATM networks using virtual finishing times

This paper is concerned with the design of a class of priority buffering strategies suitable for public, wide-area ATM networks. We specify design goals for such strategies including ease of implementation and the ability to guarantee minimum bandwidths to individual buffers. Packetized generalized processor sharing is briefly discussed and a minimum-bandwidth result for self-clocked fair queueing is given. We revisit an approach originally proposed by L. Zhang (1991) and prove that it is appropriate for ATM. Some novel, related approaches are described and analyzed.