Performance evaluation of consistency models for multicomputers with virtually shared memory

The authors introduce a deterministic and stochastic Petri net (DSPN) for evaluating the performance of an adaptive memory consistency model for a multicomputer system with virtually shared memory. DSPNs provide a unified modeling tool for both qualitative and quantitative analysis of memory consistency models. The quantitative results of the adaptive consistency model are compared with corresponding results of the sequential consistency model. The results show the limitations of the sequential consistency model and reveal the advantages of employing the adaptive consistency model.<
>