【摘要】 针对虚拟机集群资源负载不均衡的问题,提出了一种多策略与预测机制相结合的、基于虚拟机迁移的负载均衡方案.对导致系统过载的不同类型的资源采用不同的均衡策略,能够更加有效地使系统恢复正常.采用预测机制,消除瞬时峰值导致的不必要的迁移,可以减少传统的基于单阈值策略中虚拟机的迁移频率.在触发迁移后采用概率转发方式选择目的节点,能够有效解决群聚冲突问题.实验结果表明,本方案能够有效地实现系统的负载均衡.更多还原

【Abstract】 To solve the problem of unbalancing resource load status in virtual machine cluster,this paper describes a load balancing strategy which based on live migration and combines multi-strategy mechanism with the prediction mechanism.It adopts different balancing strategys for different types of resources to make the system return to normal effectively,and uses the prediction mechanism to eliminate unnecessary migration caused by instantaneous peak,so it can reduce the frequency of virtual machine migration in traditional single threshold strategy.After triggering migration,it uses probability-weighted approach to select the target node,which resolves the problem of accumulation conflicts.Experimental results show that the method can achieve load balancing.更多还原