An optimized cluster storage method for real-time big data in Internet of Things
- 10 Downloads
Data storage, especially big data storage, is a research hot spot in Internet of Things (IoT) system today. In traditional data storage methods, the fault-tolerant algorithm for data copies is adjusted with whole data file, which causes huge redundancy because there are less utilization and more cost of data storage when only a part of data blocks in the file are accessed. Therefore, an optimized cluster storage method for big data in IoT is proposed in this paper. First, weights of data blocks in each historical accessing period are calculated by temporal locality of data access, and the access frequencies of the data block in next period are predicted by the weights. Second, the hot spot of a data block is determined with a threshold which is calculated by previous data access. Meantime, in order to improve the data access efficiency and resource utilization, as well as to reduce the copy storage costs, copy of data block is dynamically adjusted and stored in different groups with high-performance and low-load nodes for data balance. Finally, experimental results show that the storage cost of proposed method is 70% less than that of traditional methods, which means that the proposed method effectively improves the data access speed, reduces storage space, and balances the storage load.
KeywordsInternet of Things Big data Real time Cluster storage Optimization
This work is supported by Programs of National Natural Science Foundation of China (No: 61502254), Program for Young Talents of Science and Technology in Universities of Inner Mongolia Autonomous Region (No. NJYT-18-B10), Teaching Reform Project Fund of University of Electronic Science and Technology of China Zhongshan Institute (No: 418YKQN02), Scientific Research Fund of Guangdong Provincial Education Department under Grant (No: 416YCQ01), Teaching Reform Project Fund of Hunan province under Grant (No: 2016), Scientific Research Fund of Hunan Provincial Education Department under Grant Nos: 16C0299, 17C0295), Natural Science Foundation of Hunan Province, China (No. 2018JJ2023). We want to thank Prof Zhang from University of Leicester for his careful checking of language to this paper.