A new method to analyze the performance of multi-access protocols

Abstract

Current telecommunications includes plenty of services and applications for final users. New standards are currently being discussed such as 802.14 for HFC networks, or DAVIC for access networks with new technologies such as HFC, ADSL, FTTC and FTTH, including new MAC multi-access protocols. Different types of traffic are defined depending on the service they are carried on: CBR, VBR, ABR and so on. Some of these services (CBR, VBR) need a deterministic access to the channel, so they need a reservation type protocol; while other services will access the channel in a random way, so they need a high-performance random multi-access protocol (RMA). Collision resolution type algorithms (CRA) are specially recommended when the network will be accessed by a large population of users in a random way. Performance of such algorithms is typically measured by the expected values of delay, throughput, and stability in the steady-state of the network. Evaluation of the performance has been rather difficult not only for the earlier algorithms such as Aloha or CSMA, but also for the most recent such as AARA, ARRA, DQRAP, etc.; difficulty is still greater when we try to evaluate new protocols proposed, as MLAP, ADAPt, etc. Where there is a combination of reservation and random access traffic. We present a systematic method to analyze multi-access algorithms using the classical queueing theory, provided that we are able to calculate the mean length of the collision resolution interval (CRT). As a result of this paper it will be possible to analyze the performance of different CRAs with a common tool and to compare them easily, even for multiple queues algorithms.