Researches On Key Technologies Of High-productivity Cloud-based Simulation Platform

With the continuous application of simulation in different domains,the scale and structure of modeled systems are getting larger and more complex.As a result,with the goal of improving on-demand collaboration and customizable modeling and simulation capabilities,the need to build high-productivity modeling and simulation platform has become increasingly prominent.As an emerging network modeling and simulation method,cloud-based simulation technology has greatly changed the modeling and simulation application mode,and provided feasible method and architecture support for building high-productivity modeling and simulation platform.However,current theory and technology researches of cloud-based simulation are still underdeveloped,and it is difficult to cope with the multi-domain multi-user on-demand collaborative customizable modeling and simulation requirements.From the perspective of productivity,the characteristics and advantages of resource integration and on-demand supply of cloud computing have not been fully utilized,and it is difficult to provide users with more agile and efficient modeling and simulation services.Aiming at the problems above and the construction of high-productivity cloud-based simulation platform,the paper proposes a user-centric cloud-based simulation serviceoriented architecture and its application mode,which solves the multi-domain and multiuser on-demand collaborative and customizable modeling and simulation requirements.Furthermore,from the aspects of modeling development,simulation execution and job scheduling,the productivity optimization framework and method for cloud-based simulation platform are proposed.The main work and contributions of the paper are summarized as follows:1.A high-productivity user-centric cloud-based simulation service-oriented architecture and corresponding application model is proposed,and a usercentric architecture for cloud-based simulation platform is designed.usercentric cloud-based simulation service-oriented architecture aims to improve the productivity of the cloud-based simulation platform.Based on the service-oriented architecture,the domain-related simulation resources,technicians and users are dynamically aggregated in a virtual collaborative modeling environment based on users' semantic requirements,which provides on-demand resource discovery,composition,collaborative modeling,and simulation in the environment.Key modules needed to implement the high-productivity user-centric simulation service-oriented architecture are designed from four aspects: user management layer,domain simulation service layer,simulation deployment and scheduling management layer and infrastructure layer.2.A productivity measuring framework for cloud-based simulation platform is proposed.Based on the concept of productivity in the field of high productivity computing,the concept and connotation of high-productivity cloud-based simulation platform are proposed,and the productivity measuring model is designed.The model quantitatively describes the key factors affecting the productivity of cloudbased simulation platform,and provides a theoretical basis for productivity optimization.3.A cloud-based domain modeling and simulation framework named Clou Fo DSim is proposed.Aiming at improving the productivity from the angles of modeling,Clou Fo DSim supports multi-domain and multi-user on-demand collaborative customizable modeling and simulation.The core of Clou Fo DSim is to build a feature-oriented domain asset tracking,management and integration method,which realizes the virtualized management of domain simulation assets under the cloudbased simulation platform and supports efficient resource discovery and on-demand dynamic aggregation.The feature-based model management and integration framework solves the problem of assets matching in classic component-based simulation.4.A bandwidth guaranteed parallel simulation job management and scheduling framework and corresponding optimization algorithms are proposed.Aiming at optimizing the productivity from the angles of job scheduling,the framework deploys parallel simulation jobs with virtual clusters to guarantee bandwidth performance between virtual machines and solve the problem of network bandwidth competition and network performance uncertainty.Ada Ba,a virtual cluster deployment algorithm with adaptive bandwidth awareness,is proposed for the allocation of virtual clusters.Furthermore,based on the Ada Ba algorithm,a bandwidth guaranteed parallel simulation job scheduling algorithm named Bg MBF and its improved algorithm named Bg MBFSDF are proposed to obtain optimal virtual resource allocation and job scheduling performance under the premise of ensuring bandwidth requirements.5.A communication-aware adaptive simulation execution optimization framework and corresponding optimization algorithm is proposed.Aiming at optimizing the productivity from the angles of simulation execution,the framework proposes efficient adaptive optimization algorithm from two aspects.One is to adaptively optimize model segmentation deployment.The paper takes the largescale artificial social simulation execution optimization as an example,and proposes a model segmentation method based on statistical movement graph partitioning,which reduces the traffic between the distributed virtual machine and meets load balance constrain.The other is to optimize the network communication consumption between virtual machines.The paper proposes a communication-aware virtual machine adaptive migration scheduling algorithm,which reduces the communication consumption between virtual machines in the data center considering the migration cost.The paper studies the high-productivity cloud-based simulation platform from two aspects: theoretical framework and key technologies.From the aspect of theoretical framework,the user-centric cloud-based simulation service-oriented architecture and its application mode for performance optimization are proposed.And productivity optimization framework is proposed from three aspects: productivity of modeling,productivity of job scheduling and productivity of simulation execution.From the aspect of key technologies,the paper proposes a cloud-based high-productivity simulation application developing method based on Clou Fo DSim.Aiming at the productivity optimization of job scheduling,a bandwidth guaranteed parallel simulation job scheduling optimization method is proposed.Aiming at the productivity optimization of simulation execution,a communication-aware adaptive simulation execution optimization method is proposed.The researches gives references and guiding for cloud-based simulation theory research and cloud-based simulation platform construction.