GRAIL : Grid-enabled Performance Analysis using Stochastic Logics

This paper aims at introducing our latest research efforts in the fields of performance analysis, stochastic logics and Grid technology adaptation. We are currently exploring a new approach for a humanfriendly and intelligent graphical specification of Quality of Service-related performance requirements for industrial-scale real-life systems. Such systems are modelled by stochastic Petri nets, on which the necessary analyses are performed. To ensure formality, expressiveness and to allow for a concise but yet unambiguous representation of potentially very complex and compound performance requirements, a new variant of continuous stochastic logics will be employed. Resulting logical formulae will be disseminated and analysed for an optimal distribution on a computational Grid back-end for the purpose of model-checking previously untractable large-scale models. The research will eventually result in a new form of stochastic logic that will feature enhanced performance requirement representation capabilities, thereby extending the depth and breadth of possible performance measurement scenarios, as well as in the development of the means to utilise for the first time the computational power of the Grid for performance analysis by distributing, parallelising and optimising required model-checking computation on a Grid cluster. In this way, available solution space will be vastly extended and truly industrial-scale systems finally become analysable.