COBOL Concurrent Access

Concurrent access requires predictability: define who writes, when, and how conflicts are handled. Favor idempotent operations when possible. This covers patterns for safe file sharing, retry mechanisms, and contention management.

Retry with Backoff

1
2
3
4
5
6
7
8
9
10
11
12
13
       01 RETRIES PIC 9 VALUE 0.
       01 MAX-RETRIES PIC 9 VALUE 5.
       01 WAIT-TIME PIC 9 VALUE 1.
       PERFORM UNTIL SUCCESS OR RETRIES > MAX-RETRIES
           ADD 1 TO RETRIES
           *> attempt I/O here
           READ CUSTOMER-FILE
           IF FILE-STATUS = '22'  *> record locked
               PERFORM WAIT-BACKOFF
           ELSE IF FILE-STATUS = '00'
               MOVE 'Y' TO SUCCESS
           END-IF
       END-PERFORM.

Implement exponential backoff to avoid thundering herds. Start with short waits and increase delay between retries. Log retry attempts for monitoring.

Serialize Critical Sections

1
2
3
4
       *> Use site standard: ENQ/DEQ, or DB locks
       PERFORM ENTER-CRITICAL
       PERFORM UPDATE-SHARED-STATE
       PERFORM LEAVE-CRITICAL.

Protect critical sections with proper locking mechanisms. Keep critical sections short to minimize contention. Always release locks in reverse order of acquisition.

Partition Workloads

1
2
3
4
5
6
7
8
       *> Partition by customer ID ranges
       IF CUSTOMER-ID >= 1 AND CUSTOMER-ID <= 100000
           PERFORM PROCESS-PARTITION-1
       ELSE IF CUSTOMER-ID >= 100001 AND CUSTOMER-ID <= 200000
           PERFORM PROCESS-PARTITION-2
       ELSE
           PERFORM PROCESS-PARTITION-3
       END-IF.

Divide work by logical partitions to reduce contention. Each job processes different data ranges, minimizing conflicts. Use consistent partitioning rules across all programs.

Monitor and Log Contention

1
2
3
4
5
6
7
8
9
       IF RETRIES > 0
           DISPLAY 'Lock contention: ' RETRIES ' retries for customer ' CUSTOMER-ID
           ADD RETRIES TO TOTAL-RETRIES
       END-IF.
       
       IF WAIT-TIME > 5
           DISPLAY 'Long wait: ' WAIT-TIME ' seconds for resource'
           ADD 1 TO LONG-WAIT-COUNT
       END-IF.

Log contention metrics for performance tuning. Track retry counts, wait times, and deadlock occurrences. Use this data to optimize locking strategies and identify bottlenecks.

• Granularity: lock the smallest safe unit to minimize blocking.
• Timeouts: detect and escape deadlocks automatically.
• Backoff: avoid thundering herds with exponential delays.
• Partitioning: divide work to reduce contention points.
• Monitoring: log contention for performance tuning.