Bandwidth management for guaranteeing transient loss performance of real-time communications

Bandwidth management for guaranteeing loss performance of real-time communications in high-speed packet-switched networks such as B-ISDN/ATM has been typically based on the concept of effective bandwidth. Although characterizing bandwidth requirements with effective bandwidth promises simplicity, effective bandwidth is derived based on some unrealistic assumptions such as the steady-state assumption, which implies that connection durations are infinite. In contrast to this assumption, connection durations in real-world networks are always finite and the transient loss behaviour of a connection may dominate the loss performance of the connection. We present a novel bandwidth management approach for guaranteeing transient loss performance of real-time communications. Our approach is simple, suitable for real-time traffic control, and does not rely on unrealistic assumptions such as infinite connection duration, infinite buffer size, or Markov traffic processes.