Reducing ATM CAC overhead by utilising aggregated traffic statistics

The quality-of-service (QoS) in ATM networks is largely determined by call admission control (CAC) and switch scheduling algorithms. A CAC algorithm determines if a requested connection can be accepted without violating QoS bounds for the new connection and for existing connections. To support QoS in ATM switches, earliest deadline first (EDF) scheduling can be implemented to ensure that the cells with the tightest time bounds are forwarded before less time critical cells are forwarded. A problem with existing CAC methods is that each node in a network needs to know the traffic parameters of all other nodes, and this requirement induces a high CAC computation and control traffic overhead. A method of aggregating connection parameters into “super connection” statistics and applying this to a CAC algorithm is developed and proved. This method reduces traffic overhead in an ATM network and enables the design of a simple EDF switch. The design of an EDF switch that uses multiple input queues, one for each super connection delay class, to sort and send time stamped cells in EDF order is developed