Research on Spatiotemporal Query Algorithms Based on Spark Streaming

With the continuous generation of large-scale mobile data, the efficient processing of spatiotemporal data has become a cutting-edge research area. To address the short-comings of existing streaming computation engines in handling spatiotemporal data, this study designs a spatiotemporal middleware that supports spatial objects, spatiotemporal partitioning, and basic spatial operations. Based on this middleware, we implement three classic spatiotemporal queries on Spark Streaming: K-Nearest Neighbors (KNN) query, Spatial Range query, and Spatial Join query. Furthermore, using a real-world dataset and a Spark cluster, we define key performance metrics—throughput, latency, and scalability to evaluate these queries in a streaming computation engine. The experimental results indicate that the throughputs of KNN, Spatial Range query, and Spatial Join query are 6912op/s, 1118op/s, and 1503op/s, respectively.