Optimization of Index-Based Method of Metadata Search for Large-Scale File Systems

With the vigorous development of the information era, the volume of data in file systems has grown to TB or EB. Metadata search determines the speed of retrieving required files in such large-scale file systems. There are many researches on metadata search based on different theories, and have made great achievements. Some researches show that the distribution of metadata presents sub-tree locality and horizontal locality, and metadata search presents heavy-tailed distribution. However, current methods do not make full use of the spatial locality and load characteristics in metadata search. In this paper, we put forward a metadata search method based on keyword-index. Our method fully exploits characteristics of metadata distribution and metadata search by reasonably partitioning indexes. The experimental result shows that our partitioning method is more efficient than current non-partitioning method.