Research On Cloud Simulation Migration Technology Bases On Performance Perception

Existing complex simulation systems based on HLA is developing in the direction of multiple spatial and temporal scales, multi-physics process, nonlinear which is strongly coupling and real configuration. Due to the complex system which is complex usage patterns, business process and increasing the requirements of capacity, communication and I/O, it's lead to the complex system is difficult to ensure the system's flexibility, robustness and fault tolerance needs during simulation, which led to the overall efficiency of the simulation system is facing huge challenges. In order to address those challenges, obtain individual needs, consistent, high-quality simulation services through cloud simulation platform which is anywhere, anytime and transparency.Existing cloud simulation platform only can allocate tasks which is according to the cloud platform load before starting the simulation tasks. It's lack of load balancing in real-time that can not to be scheduled in real-time according to the instance which is here the simulation programs running. Research on Cloud Simulation Migration Technology bases on Performance Perception for that matter. In the context of perception system performance, complete real-time scheduling, live migration tasks, and does not affect the operation of other simulation nodes. The main research results and innovation points is:1) Cloud management platform uses a modular architectureCloud management platform uses a modular architecture which consists of 3 Cores and some Extension Modules which are well-defined. Each module (or component) can be run separately, or in combination with the remaining one or more modules which is consist of the cloud simulation platform. A inversion of control technique is used to give the benefit of loose coupling and higher extensibility in application. It's make the whole platform code clean and function expansion flexibility.2) Establishing the monitoring system based on idea of'Local Monitoring-Global SynchronizationCloud simulation platform possess the characteristic of the wide area distributed and monitoring species complex. Virtualized servers as one of important part of cloud platform. The entire cloud platform as a global, virtualized servers and instance which in here as a local overall. Between cloud platform console and local monitoring for data synchronization. The performance monitoring module which in each virtualized servers that obtain performance information in real-time by VMware performance collector. The new algorithm program with improved incremental association rules can reduce the amount of data which is locally detected, it is achieve a global monitoring purposes that making data synchronization with cloud platform console. The user can use the data as needed through cloud platform console.3) To propose an improved dynamic performance prediction and evaluation based on gray clusteringTo propose an improved dynamic performance prediction and evaluation based on gray clustering by analysis the virtualized servers and virtualized machines in cloud platform which from the perspective of the overall performance of the cloud platform. Assess whether the virtual machine load reaches the threshold set which simulation Tasks located. Forecast the instance load trends to the good or bad directions when threshold is reached.4) To propose scheduling method for cloud instancesCombine with comparison, the random number method and other methods, to propose scheduling method for cloud instances. Forecast the instance load trends to the good or bad directions when threshold is reached. Making migration operations according to the proposed scheduling method. Finally, the tactical cloud simulation system applied to the consisting of cloud platforms which include above techniques and methods, cloud simulation platform composed tactics. Completion of the functional subsystems integration, to achieve the static scheduling simulation resources, the state of the global dynamic monitoring of resources, performance prediction and task-aware dynamic scheduling and migration. By testing showed that it is improved load balancing and fault tolerance simulation applications.