Transaction multicasting scheme for routing management in shared-disk database systems

Abstract

In a shared-disk environment, where computing nodes are locally coupled via high-speed network and share a common database at the disk level, it is crucial to determine at which node the incoming transactions are processed. Although a number of such conventional transaction routing algorithms have been reported, it is true that most of them are not sufficiently resilient with regard to critical system events which after all require to change the routing information. In this thesis, therefore, to afford dynamicity, we propose a new dynamic transaction routing, multicast transaction routing, MTR for short, for shared disk environments. The novel aspect of MTR is that it automatically achieves a natural balance between affinity-based routing and load balancing. When it is required to change the routing information dynamically, the routing algorithm sends multiple clones of a transaction to a group of candidate processing nodes and selects the processing node that first completes the multicasted transaction as a new processing node for re-routed transaction. The selected node is expected to be a best affinity node when the system load is evenly distributed, or a relatively unloaded processing node that is idle enough to process a transaction faster than other nodes. The simulation results indicate that MTR is capable of providing a superior performance in terms of transaction throughput. This is mainly because MTR is designed to utilize run time information of transaction processing system, such as workload distribution and buffer availability.