VegaCache: Efficient and progressive spatio-temporal data caching scheme for online geospatial applications

Abstract

With the emergence of online geospatial applications such as WebGIS services, the large spatio-temporal data and numerous concurrent users have posed grand challenges on the state-of-the-art spatial databases due to their intolerable disk I/O latency. Hence, the clients cannot be served with real-time response. In this paper, we propose VegaCache, a distributed in-memory caching scheme, to provide low-latency access for time-critical geospatial workloads. VegaCache has three significant characteristics. First of all, it is composed of tunable memory pools maintained by commodity cluster and thus its capacity can be increased linearly with more nodes. Moreover, a data replacement model is designed to determine loading hotspot data into and swapping out victim data from the distributed cache. In addition, we design a localized caching model to reduce data transfer delay with consideration of geographic proximity and geospatial access patterns. Comprehensive experimental results show that VegaCache is able to provide sustained I/O throughput and high efficiency access and thus could meet the requirements of geospatial applications.