A Novel Efficient and Robust Explicit Rate Congestion Avoidance Algorithm for ABR Service in ATM Networks

Abstract

Abstract:This paper investigates the problems in explicit rate congestion avoidance algorithms for ABR service in ATM networks. It is found that the performance of the existing algorithms [2, 3, 4] depend on network conditions which vary dramatically in the real condition such as number of traversing sessions, traffic characteristics and propagation delay. In addition, this paper proposes a novel efficient and robust ABR congestion avoidance algorithm called "Fast Rate Allocation Congestion Avoidance (FRACA)". Some simulation results are presented to illustrate performance in terms of computation accuracy and switch buffer occupancy of the proposed algorithm. Moreover, the simulation results are compared with Explicit Rate Indication Congestion Avoidance (ERICA+) and Enhanced Fast Max-Min Rate Allocation (E-FMMRA) which are the well-known explicit rate switch algorithms and also compatible with ATM Forum Traffic Management Specification 4.0. From comparison, the proposed algorithm works better than both algorithms for all configuration scenarios.

Performance analysis of congestion avoidance algorithms with non‐zero MCR guarantee for ABR service in ATM networks

“…Setting 5 ms queuing delay corresponds to 1769 ATM cells in 150:0 Mbps link. Results in Reference [8] show that the FRACA algorithm achieves the designed goals as well as the GWF ERICA+ did in Reference [7]. In order to demonstrate the enhanced performance of the FRACA algorithm in the environments that are beyond the target configurations of the GWF ERICA+, two network configurations are additionally evaluated here.…”

“…However, it is obvious that the computation steps in the GWF ERICA+ [7] are no longer similar to the ERICA+ [4]. Consequently, the problems arising may be different from the known problem in the ERICA+ [8]. In particular, the algorithm may face some limitations that actually happen in the real working environments.…”

“…FRACA algorithm applies count-based averaging interval. The timer will expire if the amount of the received ABR cells reaches the threshold, set at 1500 cells for a STM-1 link and used in the simulations (from our simulations in various network scenarios and different propagation-delay ranging from 1-2000 km; this value of averaging interval gives the best performance in term of response time and accuracy of the rate allocation [8]). In addition, the counter number may be scaled by the ratio of STM-1 speed to a given bottleneck link speed for faster or slower links.…”

“…The objective of using the 'adjusted ER' is for the queue control option. Normally, the rate of the bottlenecked session is constrained by the AR or ER max : Consequently, the switch queue length is only affected by the allocated rates of the the per averaging interval rate allocation algorithm [8]. FRACA algorithm applies count-based averaging interval.…”

“…Actually, the reason for using EAL to indicate how much the allocated bandwidth is being used by the connection is that the algorithm uses CCR field, which may not reflect the actual transmitted rate, in the rate allocation processes. Therefore in some scenarios, such as for a connection that is bottlenecked elsewhere or bottlenecked source, the allocated rate without EAL may not converge to the general weighted fairness [8].…”