Web Service Discovery for Large Scale Iot Deployments

Abstract

IoT scenarios cannot tolerate long latency in finding relevant Web services to consume on the fly or dynamically integrate in IoT applications providing real time services. In this paper, we present a comprehensive Web service discovery approach for large scale IoT deployments. We leverage the information available in the Web service description document to cluster large service repositories into functionally similar groups to reduce the discovery latency and improve precision. Then, we use statistical indexing techniques to generate data structures for each cluster for fast and efficient matching. In this research, we propose and study the performance of four matching algorithms: semantic keyword matching using Normalized Google Distance (NGD), brute force search over Term Frequency-Inverse Document Frequency (TF-IDF) matrix, K-Dimensional (K-D) tree, and Locality Sensitive Hashing (LSH). Our thesis is that indexing-based discovery algorithms (i.e., K-D tree and LSH) provide a much faster response with comparable precision, while NGD and brute force search provide a slightly better accuracy, but at the cost of high latency. Our experimental results show that we can reduce the query latency by up to 5x fold, while achieving comparable precision with the state-of-the-art service discovery mechanisms.