HFBT: an Efficient Hierarchical Fault-tolerant Method for Cloud Storage System

Ling Xiao,Beiji Zou,Chengzhang Zhu,Meng Zeng,Zhi Chen
DOI: https://doi.org/10.1109/ispa-bdcloud-socialcom-sustaincom52081.2021.00019
2021-01-01
Abstract:With the development of information technology and wide application of smart devices, data increases exponentially and more and more data are stored in the cloud, whereas most of them are infrequently accessed. Data temperature indicates the access frequency of data, where hot data is frequently accessed while cold data is rarely accessed after a while, also called archival data appearing in many fields, such as health care, finance, education etc. For these domains, appropriate fault-tolerant method for data of different temperature can improve the overall performance, so we propose the hierarchical fault-tolerant method based on temperature (HFBT) that provides the data placement strategy based on load balancing to improve application performance (The efficiency of storage and read) and recovery performance simultaneously and then executes encoding transformation according to data temperature to reduce storage overhead, and finally exploits the pipeline repair based on slice-level for archival data to decrease recovery time. This paper has three main contributions. (1) The reliability: HFBT realizes closely to 100% of the reliability. (2) The performance: HFBT saves 50% of storage overhead compared with 4-replicas and improves overall performance compared with other methods. Besides, the recovery speed of HFBT is up to 88.09% higher than that of RS (9, 6) code. (3) The flexibility: users can flexibly set the encoding transformation rules according to data characteristics.
What problem does this paper attempt to address?