BSP-Based Strongly Connected Component Algorithm in Joint Cloud Computing
Xiaochen Sun, Xingtong Ye, Kai Kang, Lijie Xu, Wei Wang, LV Lv · 2019
There is a trend that many applications are deployed among multiple public or private clouds. Due to the data asset protection and network communication cost, it is not feasible to gather all the data from different clouds to a single cloud. Inter-cloud graph data mining is becoming a new challenge. The strongly connected component algorithm is a basic graph algorithm that plays an important role in many important areas such as social network analysis, web search and even biomedical areas. However, only few distributed processing frameworks provide this algorithm, which is currently unavailable in joint cloud computing frameworks. To this end, this paper proposes a BSP (Bulk Synchronous Parallel) service over joint cloud computing and a BSP-based strongly connected component algorithm which can be easily realized on any distributed platform as long as it provides BSP service. Many graph algorithms can be easily developed in the cross-cloud computation environment by using this BSP service. Moreover, this BSP-based strongly connected component algorithm not only fills the gaps in the domain of distributed graph processing frameworks, but also extends its scalability to the level of cross-cloud computing by using the BSP service.