Design and Evaluation of a High-concurrency Web Map Tile Service Framework on a High Performance Cluster
Bo Cheng, Xuefeng Guan · International Journal of Grid and Distributed Computing · 2016
Web Map Tile Services (WMTS) have been widely used for quick and convenient sharing of geospatial information.In practice, when streaming requests to servers increase in scale, unacceptable response times and service unavailability might result.To address this scalability problem, we implemented a scalable WMTS framework on a high performance cluster (HPC), enabling the realization of elastic deployment as the client users grow in number.This scalable and high-concurrency WMTS is built totally with open-source software, including Nginx, GeoWebCache, and MongoDB.In this architecture, Nginx acts as a powerful load balancer for routing client requests; GeoWebCache is customized to publish the required WMTS and process client requests; while MongoDB is used to store the large volume of tile images in the HPC.Evaluation experiments were carried out to assess the efficiency and scalability of our WMTS system, using one synthetic workload.Experimental results illustrate that this distributed WMTS framework can achieve about 15% performance improvement when the service nodes are increased with a 0.5~2s reduction in the load time and a 5~10MB increase in network throughput.