Concurrency Control in Distributed MRA Index Structure

Answering aggregate queries like sum, count, min, max over regions containing moving objects is often needed for virtual world applications, real-time monitoring systems, etc. Since the data set is usually very large and some queries require significant processing resources, quite often such data is stored in a distributed system wherein each system handles a partition of the whole space and manages all objects in that partition. Objects keep switching from one system to another as they change their location. Currently there are no known efficient techniques for getting aggregates over moving objects while ensuring that their position updates remains atomic to the read. We introduce an efficient technique for finding aggregates over mobile objects with data stored in a distributed system by extending the multi-resolution aggregate trees to work in a distributed system and over mobile objects. We propose a multi-phase update algorithm that is highly concurrent with respect to read queries and at the same time makes modifications to the aggregate tree atomic with respect to read queries. The analysis and experimental results prove the efficacy of this update algorithm as compared to the naive update technique.