【摘要】 现有的协同检验点方法在移动环境中会带来较大的检验点过程延时 ,不能很好地支持实时事务处理 .提出了一种新的协同并行检验点方法 ,在正常的消息传输过程中 ,通过一点额外的带宽传送事务间检验点依赖关系 ;在某一事务记检验点时 ,尽可能地同时通知相关的事务记检验点 .实验表明 ,该算法对网络带宽没有明显的增加 ,而能大大降低事务记检验点的延时 ,使系统中超截止期的事务比例大大降低更多还原

【Abstract】 The existing coordinated checkpointing schemes can result in large checkpointing time latency. This paper proposes a parallel coordinated checkpointing strategy. During the normal message transportation, the checkpoint-dependency information is accompanied with the messages. When a transaction wants to take a checkpoint, the transaction tries to inform as many related transactions as possible to take their corresponding checkpoints. This can minimize the blocking time for a global checkpoint and help real-time transactions to meet their deadlines. Extensive experiments show that the proposed algorithm results in hardly any more bandwidth requirements. Furthermore, the time latency involved in a global checkpoint is reduced tremendously and there is an obvious reduction in the transaction deadline-missing rate as well.更多还原