A parallel online trajectory compression approach for supporting big data workflow

Nowadays with booming of sensor technology, location big data exhibit as high complexity, massive volume, real-time and stream-based characteristic. The current workflow systems are facing the challenge hardly to efficiently process the real-time location big data like trajectory stream. Online compression method is an available solution to preprocess these trajectory data in order to speed up the processing of big data workflow. However, the current online compression methods are in a serial execution that are hard to fast compress massive real-time original trajectory data. Aiming at this problem, we employ the multi-core and many-core approaches to accelerate a representative online trajectory compression method SQUISH-E. First a parallel version of SQUISH-E is proposed. PSQUISH-E used a data parallel scheme based on overlap technique and OpenMP to achieve the implementation over multiple-core CPUs. For further reducing compression time, we combine iteration method and GPU Hyper-Q feature to develop GPU-aided PSQUISH-E algorithm called as G-PSQUISH-E. The experimental results showed that (1) the data parallel scheme based on overlap can reach a similar SED error as the SQUISH-E (2) the proposed PSQUISH-E running on multi-core CPU achieved 3.8 times acceleration effect, and (3) G-PSQUISH-E further accelerated the effect of about 3 times compared with PSQUISH-E.