基于Remus的双机热备份优化机制研究

发布时间：2018-10-12 19:25

【摘要】：双机热备技术是虚拟化系统提高其高可用性的一种可靠的解决方案。双机热备技术一般都采用不断生成检查点的方法,将主节点上实时的状态更新传送到备份节点上面以实现主备份节点之间的状态同步。该系统在热备过程中如果遇到内存密集型应用时,传统的检查点技术会造成带宽的延迟,进而影响虚拟机双机热备的高可用性,或者在热备过程中心跳线失效而导致系统对主备虚拟机状态发生误判,从而导致系统不能正常运行。基于Remus系统的双机热备方式,提出了两种优化方案:第一是增量检查点压缩机制,第二是客户端辅助判断主虚拟机状态机制。实验测试表明,基于XOR-RLE的增量检查点压缩算法有效降低了内存密集型应用导致的带宽延迟,同时也验证了面向客户端的热备机制可大幅减少热备过程中系统误判的产生。
[Abstract]:Dual-computer hot standby technology is a reliable solution to improve the high availability of virtualization systems. In order to realize the state synchronization between the primary backup nodes, the real-time state update on the primary node is transmitted to the backup node by the method of generating checkpointing constantly. If the system encounters memory-intensive applications in the process of hot standby, the traditional checkpoint technology will lead to the delay of bandwidth, which will affect the high availability of dual-machine hot standby of virtual machine. Or failure of the jumper in the hot standby process leads to the misjudgment of the state of the main and standby virtual machine, which leads to the failure of the system. Based on the dual-machine hot standby scheme of Remus system, two optimization schemes are proposed: one is incremental checkpoint compression mechanism, the other is client aided judgment of the state of the main virtual machine. The experimental results show that the incremental checkpoint compression algorithm based on XOR-RLE can effectively reduce the bandwidth delay caused by memory-intensive applications, and it also verifies that the client-oriented hot-standby mechanism can significantly reduce the system misjudgment in the hot-standby process.
【作者单位】：华中科技大学计算机科学与技术学院;中电华通通信有限公司;
【基金】：国家高技术研究发展计划项目(2012AA012600)
【分类号】：TP309.3

【相似文献】