Adaptive techniques for BSP Time WarpReportar como inadecuado




Adaptive techniques for BSP Time Warp - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

Reference: (2002). Adaptive techniques for BSP Time Warp. DPhil. University of Oxford.Citable link to this page:

 

Adaptive techniques for BSP Time Warp

Abstract: Parallel simulation is a well developed technique for executing large and complexsimulation models in order to obtain simulation output for analysis withinan acceptable time frame. The main contribution of this thesis is the developmentof different adaptive techniques to improve the consistency, performanceand resilience of the BSP Time Warp as a general purpose parallel simulationprotocol.We first study the problem of risk hazards in the BSP Time Warp optimisticsimulation protocols. Successive refinements to the BSP Time Warp protocolare carried out to eliminate errors in simulation execution due to different riskhazards. We show that these refinements can be incorporated into the BSPTime Warp protocol with minimal performance degradation.We next propose an adaptive scheme for the BSP Time Warp algorithmthat automatically throttles the number of events to be executed per superstep.We show that the scheme, operating in a shared memory environment, canminimize computation load-imbalance and rollback overhead at the expenseof incurring higher synchronization cost.The next contribution of this thesis is the study of different techniquesfor dynamic load-balancing and process migration for Time Warp on a clusterof workstations. We propose different dynamic load-balancing algorithmsfor BSP Time Warp that seek to balance both computation workload and communicationworkload, optimizing lookaheads between processors, as well asmanage interruption from external workload.Finally, we propose an adaptive technique for BSP Time Warp that automaticallyvaries the number of processors used for parallel computation basedon the characteristics of the underlying parallel computing platform and thesimulation workload.

Type of Award:DPhil Level of Award:Doctoral Awarding Institution: University of Oxford Notes:This thesis was digitised thanks to the generosity of Dr Leonard Polonsky

Bibliographic Details

Issue Date: 2002Identifiers

Urn: uuid:783077d4-75f2-4fd2-96a9-bf1dea92711a

Source identifier: 602828856 Item Description

Type: Thesis;

Language: eng Tiny URL: td:602828856

Relationships





Autor: Unknown

Fuente: https://ora.ox.ac.uk/objects/uuid:783077d4-75f2-4fd2-96a9-bf1dea92711a



DESCARGAR PDF




Documentos relacionados