A Multi-centric Model of Resource and Capability Management in Cloud Simulation

Abstract

The popularity of the modeling and simulation application increases the demand for acquiring the service of the simulation resource and capability (R/C) anytime, anywhere, on demand through the network. Cloud simulation draws the idea of cloud computing that integrates and shares the R/Cs to provide services which significantly depending on the R/Cs' management. Most existing work assumes that there is only one management center, and some existing work uses master-slave management architecture. Both of them cannot provide services reliably in the wide area network. The relevant work in cloud computing deploys datacenters scatted and executes the distributed scheduling reliably. However, it does not take the limited R/C constraint and the parallel interoperability of the sub-tasks into account which are unique in simulation field. Our goal is to re-design an architecture of R/Cs' management with multi-centers in wide area. We also propose a mathematical model for the cross-centric global optimized allocation of R/C services. The model includes the factors of the availability of R/C service and the remote collaborative cost among sub-tasks. We show the model achieves desirable effect in the utilization balance of R/Cs and the reduction of the collaborative cost by experiment compared with an existing decentralized model.