摘要 |
PROBLEM TO BE SOLVED: To provide a parallel computer system allowing prevention of false detection of a failure even when reply return is delayed or when data transmission breaks off because a computer comes into a high load state. SOLUTION: When a failure detection part 44 detects the failure of a computer (A) 1, (B) 2 in a crossbar controller 4, a failure notification issuing part 45 refers to a Lock information storage part 43, specifies the computer (B) 2 that is a transfer opposite party of the computer (A) 1 wherein the failure occurs, and notifies the computer (B) 2 that is the transfer opposite party about the failure through a request transmission/reception part 41. A failure notification reception part 23 of the computer (B) 2 receives the failure notification through a request transmission/reception part 22, directs failure processing to a data transmission/reception part 25, and brings the computer (B) 2 into a next request receivable state. COPYRIGHT: (C)2005,JPO&NCIPI
|