Conservative DEVS
Shafagh Jafer, Gabriel Wainer
Abstract
Shafagh Jafer, Gabriel Wainer
Abstract
We present a novel conservative algorithm based on the classical Chandy-Misra-Bryant (CMB) synchronization mechanism by extending DEVS abstract simulator to provide means for look-ahead computation and null-message distribution. We integrate this mechanism into the CD++ simulation toolkit, providing a purely conservative simulator for running large-scale DEVS and Cell-DEVS models. Our algorithm is implemented on a revised DEVS abstract simulator to reduce the frequency of look-ahead computation. It also replaces time information estimations with a single lookahead computation, causing reduction in the number of null-messages. The dynamic lookahead values of the proposed algorithm are extracted from the model specification and the user is not required to provide lookahead values prior to the execution. In addition, the low-cost lookahead computation feature of the algorithm provides a fast and efficient method and reduces overhead.
OpenAlex reports 11 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 a novel conservative algorithm based on the classical Chandy-Misra-Bryant (CMB) synchronization mechanism by extending DEVS abstract simulator to provide means for look-ahead computation and null-message distribution. We integrate this mechanism into the CD++ simulation toolkit, providing a purely conservative simulator for running large-scale DEVS and Cell-DEVS models. Our algorithm is implemented on a revised DEVS abstract simulator to reduce the frequency of look-ahead computation. It also replaces time information estimations with a single lookahead computation, causing reduction in the number of null-messages. The dynamic lookahead values of the proposed algorithm are extracted from the model specification and the user is not required to provide lookahead values prior to the execution. In addition, the low-cost lookahead computation feature of the algorithm provides a fast and efficient method and reduces overhead.
Key concepts: DEVS, Computer science, Computation, Overhead (engineering), Synchronization (alternating current), Algorithm, Parallel computing, Theoretical computer science