Schedulability Analysis of Transactions in Software Transactional Memory Using Timed Automata

Software Transactional Memory (STM) is a mechanism for controlling access to shared resources in memory using an abort-restart preemption model for tasks which share data objects. This execution model of STM is different from the classical preemptive or non-preemptive models and currently the only method to determine schedulability is through an exhaustive search through the state space of all release scenarios of higher priority tasks. The existing method is costly and scales exponentially with the number of tasks making its use limited in practical situations. Timed Automata has been proven as an expressive formalism for time based systems. This paper presents a methodology for developing Timed Automata encodings for the schedulability analysis of STM systems. We validate our models using the model checker UPPAAL, and show that Timed Automata offers an efficient alternative for schedulability analysis in real-time STM.