Design of Parallel Transaction Manager on Autonomous Replicated Databases

Abstract

This paper presents a design for a parallel transaction manager in replicated database environment The parallel transaction manager preserves the autonomy of local database engine, i.e., the parallel transaction manager does not modify the transaction scheduling policies of the underlying locking based database engine. The design ensures that a user perceives database consistency as chosen by the user along with efficient processing of read queries. This paper also brings out a modification of the ABCAST communication protocol to achieve the purpose.

Indexing terms:

• Replication
• Serializability
• Atomic broadcast
• Autonomy