Virtual Network Embedding for Wireless Sensor Networks Time-Efficient QoS/QoI-Aware Approach

A recent trend in wireless sensor networks (WSNs) is network virtualization to support on-demand sharing of sensing functionality. The efficient allocation of WSN resources to sensing requests is obtained using virtual network embedding (VNE). This must take into account Quality of Service (e.g., reliability), Quality of Information (e.g., sensing accuracy), and deal with wireless interference. With increased computational complexity due to the added constraints, finding an optimal solution can be prohibitive at scale. We developed an offline embedding algorithm that searches through all possible embeddings, which allowed us to explore the tradeoff between solution quality and search time. We identify a defined set of initial processing steps that lead to high-quality solutions (within 10% of the best solution) in bounded time. We evaluated the algorithm under high stress (large networks with long paths, high data rates, beyond typical WSN configuration) to understand its limitations and the limitations imposed by the underlying WSN substrate.