HBaseSI: Multi-row Distributed Transactions with Global Strong Snapshot Isolation on Clouds
Chen Zhang, Hans De Sterck · 2011
Abstract. This paper presents the “HBaseSI ” client library, which provides global strong snapshot isolation (SI) for multi-row distributed transactions in HBase. This is the first strong SI mechanism developed for HBase. HBaseSI uses novel methods in handling distributed transactional management autonomously by individual clients. These methods greatly simplify the design of HBaseSI and can be generalized to other column-oriented stores with similar architecture as HBase. As a result of the simplicity in design, HBaseSI adds low overhead to HBase performance and directly inherits many desirable properties of HBase. HBaseSI is non-intrusive to existing HBase installations and user data, and is designed to be scalable across a large cloud in terms of data size and distribution. Key words: distributed transaction, cloud database, HBase, snapshot isolation AMS subject classifications. 68U01, 68N01, 68M01, 68P01 1. Introduction. Column-oriented data stores (column stores) are gaining attention in both academia and industry because of their architectural support for extensive data scalability as well as data access efficiency and fault tolerance on clouds. Data in typical column stores such as Google’s BigTable system [3] are organized internally as nested key-value pairs and presented externally to users as sparse tables. Each row in the sparse tables corresponds to a set of nested key-value pairs indexed by the same top level key (called “row key”). The