Restructuring the concurrent B+-tree with non-blocked search operations

Abstract

Database systems frequently establish the B-tree style indexes for fast access to data records. Since the B-tree index could be a performance bottleneck, many concurrent algorithms have been proposed to improve concurrency of B-tree accesses. In this paper we propose a new concurrent B+-tree algorithm that provides high concurrency and an efficient tree restructuring method. As the proposed method of tree restructuring always preserves a semantically consistent state of the B+-tree, a key searcher need not require any lock for a range search and a single-key search. When handling overflow or underflow in leaf nodes, it is very crucial to maintain correctly the link fields at the leaf level constructed for efficient range searches. For this, we make each leaf node contain two key-range indicators and develop a tree restructuring method using these key-range indicators.,In addition, since We prevent update processes from accessing nodes updated by others that do not commit, a recovery capability on the proposed B+-tree can be easily achieved based on the record-oriented redo/undo. (C) 2002 Elsevier Science Inc. All rights reserved.