Affinity-Based Dynamic Transaction Routing in Shared Disks Clusters

Coupling multiple nodes for high performance transaction processing has become increasingly attractive for reasons of capacity, availability and cost. The shared disks (SD) cluster couples computing nodes via a highspeed network and share a common database at the disk level. This paper proposes a transaction routing algorithm that exploits access affinities of transactions to rationalize workload on computing nodes in the SD cluster. The proposed algorithm is novel in the sense that it can achieve an optimal balance between affinitybased routing and dynamic load balancing. The simulation results show that the proposed algorithm exhibits substantial performance improvement when transaction workload is changed dynamically.