Title :
Hierarchical Composite Synchronization
Author :
Liu, Jason ; Rong, Rong
Author_Institution :
Sch. of Comput. & Inf. Sci., Florida Int. Univ., Miami, FL, USA
Abstract :
This paper presents a hierarchical composite synchronization algorithm for parallel discrete-event simulation. The composite approach combines an asynchronous CMB-style channel scanning method with a synchronous window-based method to avoid pathological situations where neither synchronization algorithms would perform optimally. The hierarchical approach addresses the discrepancy in the communication and synchronization cost for shared-memory multiprocessor multicore machines and for distributed-memory machines. The paper describes a performance model for the hierarchical composite algorithm and proposes a method for predicting the runtime empirically using linear regression. Experiment results show that the model can predict the overall performance of the hierarchical composite algorithm. A significant performance improvement has been observed over different combinations of the traditional asynchronous and synchronous approaches used separately for distributed-memory machines and on shared-memory multiprocessor multicore machines.
Keywords :
discrete event simulation; distributed shared memory systems; parallel processing; performance evaluation; regression analysis; synchronisation; asynchronous CMB-style channel scanning method; communication cost; distributed-memory machines; hierarchical composite synchronization algorithm; linear regression; overall performance prediction; parallel discrete-event simulation; performance improvement; performance model; runtime prediction; shared-memory multiprocessor multicore machines; synchronization cost; synchronous window-based method; Clocks; Computational modeling; Multicore processing; Partitioning algorithms; Prediction algorithms; Synchronization; Topology; Parallel discrete-event simulation; composite synchronization; conservative synchronization protocol; hierarchical algorithm;
Conference_Titel :
Principles of Advanced and Distributed Simulation (PADS), 2012 ACM/IEEE/SCS 26th Workshop on
Conference_Location :
Zhangjiajie
Print_ISBN :
978-1-4673-1797-9
DOI :
10.1109/PADS.2012.20