Stochastic Petri Net Analysis of Deadlock Detection Algorithms in Transaction Database Systems with Dynamic Locking
I.-R. Chen · The Computer Journal · 1995
We develop stochastic Petri net (SPN) models to analyze the best time interval between two consecutive executions of periodic deadlock detection algorithms for two-phase locking database systems with dynamic locking. Our models can accurately estimate ‘wait time per lock conflict’ automatically and allow the best time interval to be determined as a function of workload intensities and database characteristics. A system designer can apply the SPN tools developed in the paper to choose the best deadlock detection strategy, continuous or periodic, to optimize system performance, when given a set of system characteristics.