SwapQt: Cloud-based in-memory indexing of dynamic spatial data

Abstract

The ubiquity of geo-positioning technologies stimulates continuous growth in dynamic spatial datasets that fuels the development of location-based services. These services require tracking and querying a large population of moving objects. High workloads of users’ requests, both location updates and queries, need to be processed concurrently. Current solutions employ an index that is updated incrementally or rebuilt from scratch periodically. Due to the concurrency of updates and queries, current solutions still suffer from query staleness. In this paper, we present swapQt, a novel in-memory cloud-based approach for indexing dynamic spatial data that efficiently processes updates and answers queries. SwapQt consists of two main components, a routing index and local indexes. The routing index maintains the addresses of all the cloud nodes in the system and the boundaries of the data in each cloud node. Two local indexes, one to process updates and another to answer queries, are maintained and swapped periodically in each cloud node to eliminate interference between incoming updates and queries. swapQt outperformed the state-of-the-art approaches in terms of speedup and query staleness. For a workload of 1 million updates, the query staleness in swapQt was around 0.22 s compared to 4.3 s for the state-of-the-art approach. All the experiments were conducted on Microsoft Azure Cloud Computing Platform to provide realistic experimental settings.