A performance evaluation of the conservative DEVS protocol in parallel simulation of DEVS-based models
Shafagh Jafer, Gabriel Wainer
Abstract
Shafagh Jafer, Gabriel Wainer
Abstract
We present the performance evaluation of the Conservative DEVS protocol. This conservative algorithm is based on the classical Chandy-Misra-Bryant (CMB) synchronization mechanism, and extends the DEVS abstract simulator by providing means for lookahead computation and null message distribution. The protocol is integrated into the CD++ simulation toolkit, providing a conservative simulator (named CCD++) for running large-scale DEVS and Cell-DEVS models in parallel and distributed fashion. Throughout the experiments, we analyze four types of metrics, the total execution time, the average blocked time per node, the average number of positive events executed on each node, and the average number of null messages per node. We show a study on three environmental Cell-DEVS models, which shows that CCD++ provides considerable speedups, showing its ability for simulating large and complex DEVS-based models. 1.
OpenAlex reports 5 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We present the performance evaluation of the Conservative DEVS protocol. This conservative algorithm is based on the classical Chandy-Misra-Bryant (CMB) synchronization mechanism, and extends the DEVS abstract simulator by providing means for lookahead computation and null message distribution. The protocol is integrated into the CD++ simulation toolkit, providing a conservative simulator (named CCD++) for running large-scale DEVS and Cell-DEVS models in parallel and distributed fashion. Throughout the experiments, we analyze four types of metrics, the total execution time, the average blocked time per node, the average number of positive events executed on each node, and the average number of null messages per node. We show a study on three environmental Cell-DEVS models, which shows that CCD++ provides considerable speedups, showing its ability for simulating large and complex DEVS-based models. 1.
Key concepts: DEVS, Computer science, Computation, Node (physics), Synchronization (alternating current), Discrete event simulation, Protocol (science), Distributed computing