| 摘要 |
A distributed computer system includes a plurality of computer nodes, including conventional digital computer systems, mass storage subsystems, servers and the like, and a common event log. The common event log includes a plurality of storage locations for storing common event log entries. Each computer node performs processing operations in connection with a program, and generates, at selected points in its program, an event log entry including status information representing status of the computer node at the point at which the log entry was generated, the computer nodes storing the event log entries which they generate in the common event log contemporaneous with the generation thereof. As a result, the event log entries are stored in the common event log in the order in which the computer nodes reach the points in their respective programs. The common event log includes a buffer comprising a plurality of storage locations, and the location at which an entry is to be stored is pointed to by a write pointer. In one embodiment, the various computer nodes are interconnected by a common bus. When a computer node is to store a new entry in the common event log, it retrieves the write pointer, increments it and restores it in an atomic "read/modify/write" operation over the bus, and thereafter may use the write pointer which it retrieved to store the entry in the common event log.
|
