作者: Muhammad Habib ur Rehman , Sun Liew Chee , Teh Ying Wah , Ahsan Iqbal , Prem Prakash Jayaraman
DOI: 10.1109/MDM.2016.40
关键词:
摘要: The dynamic mobility and limitations in computational power, battery resources, memory availability are main bottlenecks fully harnessing mobile devices as data mining platforms. Therefore, the augmented with cloud resources edge computing (MECC) environments to seamlessly execute tasks. MECC infrastructures provide compute, network, storage services one-hop wireless distance from minimize latency communication well localized computations reduce burden on federated systems. However, when how offload computation is a hard problem. In this paper, we present an opportunistic offloading scheme efficiently tasks environments. provides suitable execution mode after analyzing amount of unprocessed data, privacy configurations, contextual information, available on-board local (memory, CPU, power). We develop application for online activity recognition evaluate proposed using event stream 5 million activities collected 12 users 15 days. experiments show significant improvement time power consumption resulting 98% reduction.