JiTS: Just-in-Time Scheduling for Real-Time Sensor Data Dissemination

Abstract

We consider the problem of real-time data dissemination in wireless sensor networks, in which data are associated with deadlines and it is desired for data to reach the sink(s) by their deadlines. To this end, existing real-time data dissemination work have developed packet scheduling schemes that prioritize packets according to their deadlines. In this paper, we first demonstrate that not only the scheduling discipline but also the routing protocol has a significant impact on the success of real-time sensor data dissemination. We show that the shortest path routing using the minimum number of hops leads to considerably better performance than Geographical Forwarding, which has often been used in existing real-time data dissemination work. We also observe that packet prioritization by itself is not enough for real-time data dissemination, since many high priority packets may simultaneously contend for network resources, deteriorating the network performance. Instead, real-time packets could be judiciously delayed to avoid severe contention as long as their deadlines can be met. Based on this observation, we propose a Just-in-Time Scheduling (JiTS) algorithm for scheduling data transmissions to alleviate the shortcomings of the existing solutions. We explore several policies for non-uniformly delaying data at different intermediate nodes to account for the higher expected contention as the packet gets closer to the sink(s). By an extensive simulation study, we demonstrate that JiTS can significantly improve the deadline miss ratio and packet drop ratio compared to existing approaches in various situations. Notably, JiTS improves the performance requiring neither lower layer support nor synchronization among the sensor nodes.