MULTICORE PARALLEL LOOSELY COUPLED FRAMEWORK FOR PROCESSING MOBILE OBJECTS
Jawdat Jamil Alshaer, Ibrahim Al‐Oqily, Sulieman Bani‐Ahmad, Adnan Alrabee · 2011
Large numbers of mobile objects and continuous queries about them characterize mobile object applications. Efficient and parallel evaluation of queries about mobile objects that continuously move is important for achieving acceptable response times. In such applications, the traditional approaches suffer from the need for parallel updates processing and real time querying and visualization. This results in poor performance in such critical systems. The emerging multicore and manycore microprocessing technologies have the potential to offer scalable performance improvement. How to explore the multicore resources to speed up mobile objects applications is thus a natural question but also a huge challenge for Moving Objects Applications (MOA). In this paper, we propose and evaluate a methodology to explore parallelism via multi-threading, transactional loosely coupled methodology design, and to implement it on multicore processors for Moving Objects Applications. We apply the proposed methodology of the loosely coupled strategy, which has been identified as the key strategy in many design optimizations for different data dependency types, to time-constrained location applications. A parallel framework and its implementation on multicore processors for the proposed strategy have been developed based on multi-threaded concepts. Experiments of using loosely coupled multithreaded parallelization to produce mobile objects parallel query processing and visualization showed clear improvement in terms of speed and efficiency.