The Impact of Network Layer on the Deadline Assignment Strategies in Distributed Real-Time Database Systems

Abstract

In recent years, Distributed Real-Time Database Systems (DRTDBS) have become one of the most exciting topics in the research of real-time computing systems. In the literature, many features of these systems have been addressed using different simulation models. However, these models always put less emphasis on the network issues. In turn, the networks are usually modeled as simple constant delay servers. However, it is found that the performance of the underlying network has a significant impact on the overall performance of the DRTDBS. The results also showed that the unpredictable behavior of the underlying network will give non-constant network transit time which in turn deter the overall system performance. To alleviate this impact, one of the methods is to assign virtual deadlines to those affected transactions such that preferences or higher priorities can be given in the scheduling of resources. In this paper, the impacts of different network issues are addressed. Then, the use of deadline assignment in transaction scheduling is discussed. Finally, a virtual deadline assignment scheme is proposed. The scheme assigns a virtual deadline based on the current network state to those transactions which require a comparatively better performance. Our results demonstrate that dynamically tuning the virtual deadlines of transactions according to the state of the network, yields a substantial improvement in the performance of that certain type of transactions without severely affecting the overall system performance