Kalman prediction method for congestion avoidance in ISDN frame-relay networks

A prediction-based congestion avoidance scheme is presented for ISDN frame-relay networks. Frame relay is a new bearer service recommended for ISDN to support packet mode communications. Unlike an X.25-based switching technique, frame-relay terminates the data link layer protocol at the end user, thus requiring flow control and error recovery to be performed end-to-end. Since the network does not send explicit congestion messages to the end users, an effective congestion avoidance scheme is needed to prevent severely degraded performance which can result from network congestion. A prediction method for the congestion avoidance scheme is presented. It is based on the recursive digital filter known as the Kalman one-step predictor. The model developed from this idea is analyzed briefly to demonstrate adaptive end-to-end window flow control and error recovery procedures, which depend upon the predicted congestion states of the network