The Research Of Task Scheduling In Distributed Real-Time Database

With the rapid development of IT technology, the developing process of the traditional industry speed up rapidly. Intelligent security monitoring is one important aspect of it. Because of the need for collecting massive real-time data and historical data which have been gradually evolving into hierarchy and distribution, and dealing with them within the limited time, distributed real-time database used to replace traditional centralized relational database has become an inevitable trend.But how to assign the sub-transactions which are decomposed by transaction to each parallel processing node, to ensure the integrity, synchronization and security for transaction executing are the questions which must to be solved. In fact, many sub-transactions are not the same sizes which content the different operations. So the sub-transaction which contents a lot of operations in the busy node is hard to schedule to other idle nodes to make system load-balance. The paper bases on the heterogeneous distributed real-time system and the sub-transaction can be further decomposed, to make each node in the system load balance and reduce the total time-consuming of transaction processing.In each layer of the system, the priority of sub-transaction is determined by the end time of precursors, transmission losses and the earliest time the node can execute. Then system schedules the transaction to the appropriate node. If nodes do not load balance, the system replicates the sub-transactions which are decomposed by the local transaction scheduled from master host to the other idle nodes. Because the local transaction can be decomposed, so the result of the load-balancing schedule will be more perfect. The distributed real-time database will perform well to handle the complex parallel applications. Finally, experiment verifies the improvement strategies dealing with complex transactions in performance.