Distributed multiprocess transaction processing system and method

A method for fault tolerant transaction processing comprising the steps of

• establishing an association between each of a plurality of processing units and at least one non-volatile storage media having at least two ports, whereby only one processing unit is allowed to address the associated at least one non-volatile storage media so long as the processing unit continues to operate and does not relinquish control of the at least one non-volatile storage media,
• storing information concerning status of transactions coordinated by each processing unit in a log on the associated at least one non-volatile storage media,
• interconnecting the plurality of processing units through at least one interprocessor communications link,
• monitoring signals transmitted from time to time from each processing unit to determine whether any of the processing units has failed,
• permitting a second processing unit to establish communications with the at least one non-volatile storage media formerly associated with a failed processing unit and to examine the transaction information for the transactions formerly coordinated by the failed processing unit, and
• causing the second processing unit to complete the transactions formerly coordinated by the failed processing unit by committing or aborting those transactions.

A method for fault tolerant transaction processing in a processing system having a plurality of processing units, non-volatile memory means for storing a log therein for each processing unit, and interprocessor communications links connecting each of the processing units, the method comprising the steps of

• creating a log in the non-volatile memory means for each of the processing units;
• establishing a transaction coordinator in one of the processing units for each transaction processed by the system,
• recording information developed by the associated transaction coordinator about the associated transaction in the log,
• detecting the failure of the processing unit wherein the transaction coordinator resides,
• causing a second processing unit to establish communications with the log associated with the failed processing unit,
• scanning the log to determine entries which are potentially inconsistent with information stored elsewhere in the system,
• interrogating other processing units to identify inconsistencies between the log and information stored elsewhere in the system, and
• committing and aborting transactions in accordance with the actual state of those transactions as necessary to cause the log and the information stored elsewhere to be consistent.

Fault tolerant transaction processing apparatus comprising:

• means coupling each of a plurality of processing units and one non-volatile storage media associated therewith to enable only one processing unit to address the associated at least one non-volatile storage media so long as the processing unit associated therewith continues to operate and does not relinquish control of such at least one non-volatile storage media,
• each of said non-volatile storage media storing thereon information concerning status of transactions coordinated by the associated processing unit,
• means including an interprocessor communication link interconnecting the plurality of processing units,
• means responsive to signals transmitted from each processing unit to determine whether any of the processing units has failed for enabling a second of the plurality of processing units to establish communications with the at least one non-volatile storage media formerly associated with a failed one of the plurality of processing units, and to examine the information stored on such formerly-associated non-volatile storage media concerning status of transactions formerly coordinated by the failed one of the plurality of processing units, and
• said second processing unit thereby being enabled to coordinate the transactions formerly coordinated by the failed one of said plurality of processing units for completing or aborting those transactions.

Fault tolerant transaction processing apparatus including a plurality of processing units and a plurality of non-volatile memory means and including interprocessor communications links coupling each of the plurality of processors, the apparatus comprising:

• means operatively coupling each of the processing units an associated non-volatile memory means for storing thereon a log for the processing unit associated therewith;
• at least one of the plurality of processing units including transaction coordinating means for each processed transaction;
• means for recording in the log of a non-volatile memory information developed by the associated transaction coordinating means about the processed transactions;
• means for detecting the failure of a processing unit which the transaction coordinating means for enabling a second one of the plurality of processing units to establish communications with the non-volatile memory means having there the log associated with the failed processing unit;
• a file system for storing information about the actual states the processed transactions;
• means for scanning the log associated with a processing unit to determine entries of information which are inconsistent with information stored in the file system;
• and
• means for interrogating other processing units to identify inconsistencies between the logs associated therewith and information stored in said file system for enabling a processing unit to commit or abort transactions in accordance with the actual state of those transactions as determined by information in the associated log being consistent with information stored in the file system.

A method for coordinating fault tolerant transaction processing within a system having a plurality of processing units, an operating system, and a plurality of transaction coordination logs stored within non-volatile memory means, and including interprocessor communications links connecting each of the processing units, the method comprising the stops of:

• creating a transaction manager within the operating system for each transaction coordination log;
• establishing a transaction manager for each transaction processed by the system;
• recording information developed by the transaction manager about a transaction in a transaction coordination log;
• detecting a failure of the processing unit on which a transaction manager is executing during a transaction;
• creating a backup transaction manager within the operating system for another processing unit;
• establishing communications between the backup transaction manager and the transaction coordination log associated with the failed transaction manager;
• scanning the transaction coordination log to determine entries are potentially inconsistent with information stored elsewhere in the system;
• interrogating other Processing units to identify inconsistencies between the transaction coordination log and information stored elsewhere in the system, and
• committing and aborting transactions in accordance with the actual state of those transactions as necessary to cause the transaction coordination log and the information stored elsewhere to be consistent.