中文摘要：

现有的虚拟机池化管理普遍采用的是master/slave模式，存在单点失效问题。针对此问题，设计并实现一种基于角色的虚拟机池单点失效处理方法。该方法采用角色划分和选取机制自动指定主备节点，并基于序列法解决了自动选取机制中存在多个master节点冲突问题。实验表明，该方法能够实现master节点对用户透明的失效恢复。

英文摘要：

The master/slave mode is the most pervasive management policy for virtual machine resource pool at present, but has the problem of the single point of failure. In light of this, we design and implement a role-based method to solve the single point of failure for virtual machine resource pool. In this method, the master or backup node can be automatically activated according to the role appointment and the selection mechanism. Furthermore, a sequence method is proposed to solve the conflicts among multiple master nodes in automatic selection mechanism. Experiments show that the method can realise the recovery against the failure of transparency of master node on users.