A relevance and redundancy aware performance feature selection approach for distributed file system

作者：Zhao Tiezhu^*; Yuan Huaqiang

来源：Journal of Computational Information Systems, 2013, 9(10): 4011-4018.

DOI：10.12733/jcis5907

• Distributed file systems
• Performance factors
• Performance measurements
• Performance metrics
• Performance prediction
• Redundant features
• Relevance and redundancies
• Selection methods

Obtaining good performance measurements is a critical first step in effectively modeling and predicting the performance of distributed file system. In this paper, we are interested in studying the performance factor selection methods of distributed file systems. A relevance and redundancy aware performance feature selection approach is proposed for selecting a subset of most informative performance factors. We measure the accuracy of performance prediction based our proposed approach, and analyze the accuracy of different performance metrics, namely bandwidth, throughput and latency. The result indicates that our approach could effectively remove irrelevant performance factors, eliminates redundant features and provides better accuracy in distributed file system.

• 出版日期2013
• 单位东莞理工学院