Virtual machine fault rapid recovery method and system

Abstract

The invention discloses a virtual machine fault rapid recovery method and system. A virtual machine address mapping table is backed up, a host machine physical page corresponding to the virtual machine is set to be read-only, protection of host machine physical memory content corresponding to the virtual machine is achieved through the write-time copying technology, and after the virtual machine breaks down, the previous backup state of the virtual machine can be rapidly recovered. The method has the advantages of being small in backup data size, high in backup and recovery speed and the like,and reliable operation of the virtual machine is guaranteed.

Description

technical field [0001] The invention relates to a method and system for quickly recovering from a virtual machine fault, and belongs to the technical field of virtual machine fault processing. Background technique [0002] At present, the main method to improve the reliability of virtual machines is checkpoint technology. Checkpoint technology is a mirror file that saves the instantaneous state information of a process or system during operation. According to certain needs, the system can be rolled back to a previous running state status. The virtual machine memory checkpoint technology mainly has four important quantitative indicators, which are checkpoint size, downtime, total save time, and recovery time. [0003] Memory exclusion is an important means to reduce the size of the checkpoint. The part of the memory that does not need to be saved in the process includes read-only and dead. The incremental mechanism is a well-known checkpoint optimization technique, which on...

AI Technical Summary

Problems solved by technology

This mechanism takes advantage of the existing features of Xen and simplifies the implementation of virtual machine checkpoints, but the checkpoints implemented by this method are not suitable for time-sensitive applications, such as system Debug or high-reliability systems that require accurate output state protection, because migration is a process that takes a long time, the migration synchronization mechanism based on the pre-copy algorithm makes the final saved checkpoint state always ahead of the time point that should be captured (that is, the moment when the checkpoint is triggered)

However, there is a major defect in this method. Since the migration of the virtual machine takes a certain amount of time, the triggering moment of the migration always lags behind the moment when the virtual machine is stopped and the state of the virtual CPU is saved. Therefore, the checkpoint file saved by using the virtual machine online migration technology The state of the virtual machine is always ahead of the state of the virtual machine when the system triggers the save checkpoint

That is to say, when the instantaneous state of the system at time t is needed, the system state at time t+Δt (Δt is the total time of virtual machine migration) is saved by using this strategy. unacceptable

Images