A new admission control policy and over load detection in DRTDBS

Abstract

Today's Real Time System (RTS) is characterized by managing large volume of dispersed data making real-time distributed data processing a reality [1,2,14]. However, such a rapid growth of distributed real time database application executes in unpredictable environment where workload varies significantly at run time. Hence, computation workload on a large scale distributed system can lead to numbers of transaction deadline misses [3,4]. Therefore, efficient database management algorithm and protocol for accessing and manipulating data are required to satisfy timing constraints for supported application. In this paper, an algorithm has been proposed for admission control policy and overload detection for Distributed Real Time Database Systems (DRTDBS) consisting of local controller and global load balancer working at each site, which decide whether to admit or reject the newly arrived transaction. The local controller controls the admission process of incoming transaction whereas the global load balancer collects the performance data from other nodes and balances the system wide workload. An outright rejection of a transaction may be safer than attempting to execute that transaction only to miss its deadline. The simulation results show that the new algorithm successfully balances the workload in a distributed real time database and provide lesser transaction miss ratio under tight transaction workload as compared to existing protocols [5-7].