Improving Throughout of Continuous k-Nearest Neighbor Queries with Multi-threaded Techniques
Traditional moving objects database has faced the rapid evolution of modern CMP processor. To evaluate massive concurrent continuous queries towards moving objects, parallel processing techniques and cacheconscious algorithms adapting to memory hierarchy and multi-core architecture should be developed to maximize the processor computation abilities. This paper introduces a multi-staged engine (MSE) for high performance and adaptable execution of massive concurrent continuous queries processing, which exploits pipeline strategy and departs the continuous query processing into three simultaneous stages: preprocessing, executing and dispatching modules to improve the parallelism with multi-threaded technology. Based on MSE framework and grid index for moving objects, we present a multi-threaded algorithm (MT-CNN) for massive continuous k nearest neighbor queries processing. MT-CNN algorithm uses threaded workload parallelism and cache-conscious execution reorganization strategies to improve the spatial and temporal locality. Experimental evaluation on a dualcore platform and analysis show that MT-CNN algorithm achieves a performance improvement over the existing traditional optimization counterparts.
CKNN queries multi-thread MSE framework MT-CNN algorithm pipeline strategy
LIAO Wei WU Xiao-Ping ZHANG Qi ZHONG Zhi-Nong
School of Electronic Engineering Naval University of Engineering Wuhan,China College of Electronic Science and Engineering National University of Defense Technology Changsha,Chi
国际会议
上海
英文
2253-2257
2009-11-20（万方平台首次上网日期，不代表论文的发表时间）