Study Of Extended T-Tree Index Mechanism For Business Object

The emphasis of this paper is on index mechanism for main memory business objects in the context of performance critical application servers.In order to decrease the layer of tree, consequently decrease the cost of disk I/O, the index structure in traditional DRDBs (Data Resident Databases) is usually B-tree or the variables of B-tree. For the very fast speed of main memery access, this paper consults the index approach in MMDBs (Main Memory Databases) , and adopts a novel index structure - T-tree, in the context of organization and management of main memory business objects.Based on a detailed analysis of typical T-tree, an improved schema, T-tail tree, is presented, which decreases the rotation times, hence improves the index performance, specially in concurrent access to T-tree. Then the modification of index item is made, sustaining the business objects including variable length, multi-attributes and repeated keywords.Finally, a T-tree concurrent control protocol is proposed, which is based on logical versioning and physical versioning. The major characteristics are that read-only transactions can run without any locks and splitting the difference between improving the concurret degree and decreasing the cost of lock and unlock.