Hash-Based Multi-Attribute Database Indexing on The Cloud

Recent years, the Cloud platform provides an ease to use interface between providers and users, allow providers to develop and provide software and databases for users over locations. Currently, there are many Cloud platform providers support large-scale database services. However, most of these Cloud platform architectures only support simple keyword-based queries and can’t response complex query efficiently due to lack of efficient in multi-attribute index techniques. The existing multi-attribute index structures for Cloud platform are based on traditional R-tree, k-d tree and Quad tree, but there is still without study for evaluating these schemes yet. Moreover, there has an assumption that the data on Cloud platform is distributed into local slave nodes by range distribution. Such that a sequence of value intervals of attributes in a local slave node can be denoted as a node cube. These node cubes are maintained in the global index of master nodes for pruning irrelevant data and indicating the slave nodes for searching. However, range distributed data may cause load imbalanced in slave nodes due to data may massed in some small range by the property of normal distribution. To address this problem, hash distribution is usually applied. In this paper, we propose a new multi-attribute index structure, which combines hash-based scheme and tree indexing, for Cloud platform to manage the huge and variety data. Our experimental results demonstrate that our proposed index structure outperforms existing tree-based only indexing.