COOPERATIVE DISTRIBUTED SYSTEM, AND JOURNAL AND RECOVERY PROCESSINGS THEREIN

摘要

The transaction processing throughput of a system in which a plurality of servers cooperatively perform distributed processing of transactions is improved. The servers (1a, 1b, 1c, ...) are connected to each other through a communication network (2), and a resources (16) and a log 18 are distributed over these servers. When one of the servers, e.g., the server (1a) receives a transaction, the server (1a) instructs the servers (1b, 1c, ... ) over which the resource 16 relating to this transaction are distributed to update their resources, acquires the log data representing the updating contents and stores the log data in its log (18) . The servers (1b, 1c, ...) update the resources in accordance with the instruction. Even when a fault occurs in one of the servers (1b, 1c, ...) during the transaction processing so that the updating of the resources has not been executed, the system assumes that all the resource has been updated and proceeds to processing the following transaction. The server in which the fault acquires the log data related to the failed transaction from the log (18) of the server (1a) after the server has restored from the fault, and correctly updates its own database (16) based on the log data. The log data storing method for the transaction processing, the log retrieving method for the recovery processing, and the method of judging whether or not the resources have been updated during the recovery processing are all improved and of high speed.