Optimistic Concurrency Control for Multi-hop Wireless Sensor Networks

With the inclusion of actuation capabilities, emerging wireless sensor applications are much less tolerant to inconsistencies in decisions compared to passive sensing applications. Multi-hop networks suffer from these problems more profoundly as they cannot directly utilize atomic broadcast operations for coordination. In this study, we provide a lightweight single hop primitive, Read-All-Write-Self (RAWS), that achieves optimistic concurrency control. RAWS guarantees serializability, which simplifies implementation and verification of distributed algorithms, compared to the low level message passing model. We also present a self-stabilizing multi-hop extension of RAWS, called Multi-hop Optimistic Concurrency Control Algorithm (MOCCA), to address the challenges of multi-hop networks. MOCCA improves the performance of RAWS transactions in multi-hop networks while maintaining serializability. We implement MOCCA in JProwler simulator using TDMAand CSMA-based MAC layers and compare it against a lightweight locking scheme and serial execution of transactions. Our results indicate that concurrent execution in MOCCA can outperform serial execution in task completion time via better utilization of broadcasts and concurrent execution. Finally, we also show that under heavy loads optimistic concurrency control provides much better performance compared to locking schemes.