Adaptive message scheduling for supporting causal ordering in wide-area group communications

Abstract

Most of existing causal ordering protocols employ the passive message ordering (PMO) approach, which passively re-orders the disturbed messages at destinations. Since the PMO approach cannot prevent disturbance of message ordering, seriously disturbed messages can overrun the receiving buffer of these protocols and cause damaging message loss. This paper proposes the adaptive message scheduling (AMS) approach, which introduces scheduling latency for messages at sources to alleviate the disturbance of message ordering at destinations. We develop a theoretical base for calculating the value of scheduling latency. The scheduling latency can adapt to varying channel latency and is related to the message ordering relation and the message lifetime. We then describe a new protocol that employs the AMS approach. On basis of a proposed model for evaluating group communication protocols, we conduct experiments and simulations to validate the AMS approach. The simulation results show that the AMS approach can effectively improve the ordering latency and reduce the receiving buffer size. A small penalty of employing the AMS approach is a larger loss ratio if strict real-time delivery of messages is to be guaranteed.