Dynamic Multiversion Database System Based On B～(link)-tree Index

Today.database management systems are widely used in many application fields and greatly affect the efficient running of these applications. Both concurrency level and transaction throughput are crucial elements which affect database performance. So how to improve database performance and concurrency has always been one of hotspots in database research.Multiversion storage technology may efficiently improve the database concurrency and transaction throughput. In the multiversion schema,a transaction can get the correct version of a data item without holding and locks or latchs and the read-only transactions can proceed without being blocked by any update transaction.The B^link-tree ,a variant of the B-tree,has all of the nodes at a particular level chained together into a linked list and has super-high concurrency. When searching a value in the tree, a read operation never locks any node and the splitting of a node that a update operation may necessinate has no effect to the read operation.so to increase database concurrency, it is a good idea to choose B^link-tree as database index.This paper provides a new scheme that combines B^link-tree-based database index with multiversion technology. In this algorithm the paper implements the field-level lock that is very small granularity lock.Test results show that this scheme can improve the database performance and trarsaction throughput in concurrent environment significantly.Because the mysql is an open source software and has good and easy extensible interfilces,this paper implements the algorithm in mysql database management system and test it with super-smack test tool and compare it with traditional B-tree index algorithm,and provides the test results.