Enhancing File Recovery from Distributed File Systems (DFSs) Using Erasure Coding and Replication
Prince Raj, Sapna Sinha · 2024
Several solutions have been used in distributed file systems to increase data availability and dependability. Historically, replication techniques were the favored method for storing data in such systems. However, there has been a recent movement towards the use of erasure-coding (EC) methods to overcome space efficiency concerns. While EC is more successful than replication approaches in overcoming space efficiency difficulties, it introduces its own performance degradation factors, such as encoding and decoding delays, as well as input and output (I/O). To address these issues, this research proposes a buffering and combining strategy. This solution combines several I/O requests made throughout the encoding process in an EC-based distributed file system into a single operation, resulting in more efficient processing. In addition, the research recommends four recovery strategies to reduce the impact of decoding on disk input/output loads. These measures include disk input/output load distribution, random block layout, multi-thread parallel recovery, and the matrix recycling approach. All of these methods are intended to optimize the distribution of disk input/output loads produced during the decoding process.