An Access Prefetching Strategy for Accessing Small Files Based on Swift

Abstract

With the burst of Internet data, the real-time storage efficiency of massive small files has become a challenge for the cloud storage industry. Compared with large files, real-time access to massive small files can put tremendous pressure on file systems. Most distributed storage systems focus on large files in terms of network communication, metadata access and data layout, which affect the IOPS performance of massive small files. In this paper, aiming at the shortcomings of current distributed storage systems in file relevancy assessment, this paper proposes a new object relevance assessment method based on the relevance measure of object access timing and attribute semantics. We firstly introduced the present research on data prefetching and relevance analyzing. Then, we define files relevance and analyze the influencing factors on relevance of files. File prefetching model is constructed by assessing the relevance of the files. At last, the validity of the proposed strategy is verified through experimental tests.