To improve fairness in DQDB metropolitan area networks a mechanism known as bandwidth balancing (BWB) has been introduced in the final version of the IEEE 802.6 standard document. Since then several researchers have proposed more or less significant changes to the basic mechanism, attempting to adhere to different interpretations of the fairness concept, and to match the needs of different categories of users. In this paper attention is focused on the adaptive class of fairness control mechanisms and two solutions are analyzed and compared: the first one allows for a fair sharing of the bandwidth by keeping the control rate of each node proportional to its offered traffic, whereas the second one keeps the control rate proportional to the number of segments queued in each node waiting for transmission. In order to do this, an analytical model suited to study a DQDB network under overload conditions is introduced. The model is also amenable to the evaluation of the transient behaviour of DQDB networks, thus enabling a thorough investigation of the performance that can be obtained by adopting different fairness control mechanisms. The analytical model has been validated through the use of a software simulator. Numerical indices measuring the performance of the balancing mechanisms mentioned above are also given
Comparison of Adaptive Fairness Control Mechanisms for DQDB Metropolitan Area Networks
G Cena;L Durante;A Valenzano
1995
Abstract
To improve fairness in DQDB metropolitan area networks a mechanism known as bandwidth balancing (BWB) has been introduced in the final version of the IEEE 802.6 standard document. Since then several researchers have proposed more or less significant changes to the basic mechanism, attempting to adhere to different interpretations of the fairness concept, and to match the needs of different categories of users. In this paper attention is focused on the adaptive class of fairness control mechanisms and two solutions are analyzed and compared: the first one allows for a fair sharing of the bandwidth by keeping the control rate of each node proportional to its offered traffic, whereas the second one keeps the control rate proportional to the number of segments queued in each node waiting for transmission. In order to do this, an analytical model suited to study a DQDB network under overload conditions is introduced. The model is also amenable to the evaluation of the transient behaviour of DQDB networks, thus enabling a thorough investigation of the performance that can be obtained by adopting different fairness control mechanisms. The analytical model has been validated through the use of a software simulator. Numerical indices measuring the performance of the balancing mechanisms mentioned above are also givenI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.