大数据环境下的可靠存储技术思考

[摘要] 针对分布式容错技术的研究，提出了两点关键要求：降低冗余开销、提高节点修复效率。分析目前主流的容错策略: 复制、纠删码、再生码、基于局部可修复码，并认为这些容错策略存在不同程度的缺陷，因此设计出容错能力、计算效率及存储利用率更高的容错策略，仍是未来很长一段时间内值得深入研究的问题。

[关键词] 大数据；可靠性；分布式存储；容错技术

[Abstract] Two key requirements of fault tolerance technology are proposed in this paper: minimal storage overhead and maximum node recovery performance. Four main strategies for fault tolerance are analyzed: replication, erasure codes, regenerating codes and locally repairable codes. It is considered that these fault tolerance strategies have different defects. Designing a fault tolerance strategy with higher fault tolerance, better computational efficiency and memory utilization will still be a problem needs to be solved in the future.

[Keywords] big data; reliability; distributed storage; fault tolerance technology