A partitioned B-tree is an ordinary B-tree with an artificial leading key attribute added. More…

A partitioned B-tree is an ordinary B-tree with an artificial leading key attribute added. More specifically, if X is the key of the relation to be indexed, the partitioned B-tree uses the key .(P; X ), where the leading key attribute P takes partition numbers as values. Partition 0 is the regular partition in which most of the tuples usually reside. When a new bulk of tuples is to be inserted, a new partition number i>0 is reserved and the tuples .x; v/ from the bulk are inserted into the B-tree as tuples .i(x; v). This insertion is executed as part of the inserting user transaction.

View complete question »A partitioned B-tree is an ordinary B-tree with an artificial leading key attribute added. More specifically, if X is the key of the relation to be indexed, the partitioned B-tree uses the key .(P; X ), where the leading key attribute P takes partition numbers as values. Partition 0 is the regular partition in which most of the tuples usually reside. When a new bulk of tuples is to be inserted, a new partition number i>0 is reserved and the tuples .x; v/ from the bulk are inserted into the B-tree as tuples .i(x; v). This insertion is executed as part of the inserting user transaction. Later when the transaction has committed, tuples in partitions i>0 may gradually be migrated to partition 0, using system transactions (redo-only structure modifications). Conversely, when a bulk of tuples is to be deleted, system transactions may first be run to migrate the tuples to a new partition, which is then deleted as part of a normal user transaction. What are the obvious advantages and disadvantages of the partitioned B-tree for bulk actions and single-tuple actions? Give a multi-granular locking protocol for bulk-read, bulk-insert, and bulk-delete actions on the partitioned B-tree.

View less »