Traffic Engineering in ATM Networks: A Dynamic Leaky Bucket-Based Approach

摘要

In this paper we present a new state estimation method based on a Dynamic Leaky Bucket (DLB) mechanism. The DLB provides information about the instantaneous bit rate of a source allowing more effective flow control and can achieve a good match in terms of predicting the Congestion State of the switch. We show that the DLB is able to police implicit resource management which is used for both Constant Bit Rate (CBR) and Variable Bit Rate (VBR) traffics, and explicit resource management which is used for Available Bit Rate (ABR) traffics and Unspecified Bit Rate (UBR) traffics. The DLB policy is based on large deviation theory that includes the identification of appropriate energy, entropy function and effective bandwidth of a source.