Bf-Matrix: A Secondary Index For The Cloud Storage

Although people have proposed many kinds of NoSQL databases, also referred as Key-Value stores, there is still lack of an efficient solution for the problem of non-key attribute queries. In this paper, we propose BF-Matrix, a hierarchical index composed of bloom filter and B+ tree. Faced with the massive data and the large scale cluster, the layered solution could shorten the search path and make the best of scattered resources. Moreover, it is able to scale up and scale back according to the changes of data size and cluster scale, and isolate the job of update and retrieval in a limited scope. To eliminate the risk of false negative and to ensure our index "look like consistent", two rules are given to specify the behavior of index update and data retrieval. Experimental results demonstrate that our solution not only outperforms the state of the art, but also is flexible enough to adapt to the cloud environment.