发明名称 PARALLEL COMPUTING SYSTEM AND MIGRATION METHOD
摘要 A first node stores therein accessed-area information that indicates a first area included in a storage area assigned to a target process, the first area having been accessed by the target process for a latest predetermined period of time. The first node stops the target process when an instruction to migrate the target process is received, transmits, to a second node, state information that indicates a state of the target process when the target process is stopped, and transmits first data stored in a first area to the second node, on the basis of the accessed-area information. The second node receives the state information and the first data from the first node, generates the target process on the basis of the state information, and restarts the target process using the first data.
申请公布号 US2017041191(A1) 申请公布日期 2017.02.09
申请号 US201615202034 申请日期 2016.07.05
申请人 FUJITSU LIMITED 发明人 KATO Takeharu
分类号 H04L12/24;H04L29/08;H04L12/707 主分类号 H04L12/24
代理机构 代理人
主权项 1. A parallel computing system that migrates, to a second node, a target process that is being executed in a first node, the first node and the second node being one of the nodes that are an execution unit for processing of executing a process, wherein the first node comprises: a storage configured to store therein accessed-area information that indicates a first area included in a storage area assigned to the target process, the first area having been accessed by the target process for a latest predetermined period of time;a processor in the first node configured to stop the target process when an instruction to migrate the target process is received; anda transmitter configured to transmit state information that indicates a state of the target process when the target process is stopped, and to transmit first data stored in the first area, on the basis of the accessed-area information, and the second node comprises: a receiver configured to receive the state information and the first data; anda processor in the second node configured to generate the target process on the basis of the state information and to restart the target process using the first data.
地址 Kawasaki-shi JP