A Bottom-Up Tree Based Storage Approach for Efficient IoT Data Analytics in Cloud Systems

Internet of Things (IoT) has been widely applied in various domains, e.g. environmental monitoring, intelligent transport system, video surveillance, etc. In most of the IoT applications, the IoT data is generated from a number of data sources, not just only one source. In addition, IoT data has various types with different processing requirements. The high-priority IoT data should have better storage and processing manners than the low-priority IoT data. The objective of this paper is to propose an efficient cloud storage approach for considering the multi-aspect requirements of IoT data. In the approach, a light-weight data structure is used to depict the distribution and calculate the size of each IoT subset (type) in all data sources. Then, we form a number of storage-locality groups from cloud storage blocks. However, the storage-locality groups have different storage sizes and locality capabilities. We would like to place the high-priority IoT subset in the storage-locality group with a strong locality capability. Therefore, there is the placement-combinational problem between IoT subsets and the storage-locality groups. To efficiently solve the IoT placement problem, we propose a bottom-up tree based approach associated with the solution of the well-known combinatorial problem: knapsack. Considering the knapsack problem with the NP-hard computational complexity, we also propose a heuristic placement approach.