Real-time Performance Monitoring

Once AUTOGEN has adjusted the basic system parameters, other programs can be applied if the manager believes there are performance problems. These include the following:

• Examining the error log for hardware problems, discussed earlier

• Using DECevent to determine if unusual events are occuring

• Examining system performance statistics via MONITOR and the particular TCP/IP stack the manager has installed

• If DECnet or TCPIP is used, examining network performance statistics

• Tracking other resources with accounting, such as printer usage

• Using SHOW CLUSTER to track cluster resources

System performance can be monitored in real time or recorded for future analysis by the MONITOR program. This task is described in the Open VMS System Management Utilities Reference Manual. This utility has about 20 suboptions, and data can be presented in a graphic-like display (the first example that follows) or a tabular display (as in the second example).

• MONITOR SYSTEM summarizes several of the other options. The display looks like this and is updated every six seconds:

   Node: NODE1            OpenVMS Monitor Utility    12-FEB-2002 12:44:32 Statistic: CURRENT        SYSTEM STATISTICS                                                       Process States        + CPU Busy (100)          -+        LEF:    25      LEFO:     0         | *************************|        HIB:    11      HIBO:     0 CPU  0 +--------------------------+ 100    COM:     8      COMO:     0        |*                         |        PFW:     0      Other:    0        +--------------------------+        MWAIT:   0        Cur Top: BATCH_77 (6)                         Total: 44        + Page Fault Rate (1468)   -    + Free List Size (35173)    +       |****|*********************|   |****************            |54K MEM  0 +--------------------------+500 + --------------------------+        |****                           |*************              |5765        +--------------------------+    + Mod List Size (3078)      +        Cur Top: BATCH_29 (78)        + Direct I/O Rate (412)   -+    + Buffered I/O Rate (110)  -+        |**********************    |    |*****                      | I/O  0 +--------------------------+500 +---------------------------+ 500       |*                              |                           |        +--------------------------+    +---------------------------+        Cur Top: BATCH_77 (23)          Cur Top: BATCH_77 (6) 

• MONITOR IO /ALL produces a display similar to the following. I/O on a specific device can also be monitored.

                         OpenVMS Monitor Utility                         I/O SYSTEM STATISTICS                             on node NODE1                          10-FEB-2002 22:12:44                            CUR        AVE       MIN        MAX     Direct I/O Rate      14.33       4.46      0.33      15.33     Buffered I/O Rate    20.91      47.47     24.91      69.00     Mailbox Write Rate    0.00       0.45      0.00       2.95     Split Transfer Rate   1.66       1.56      0.33       3.97     File Open Rate        1.66       1.26      0.33       2.98     Page Fault Rate      20.58      52.31     17.33     178.00     Page Read Rate       14.29       9.00      0.00      26.88     Page Read I/O Rate    2.65       2.43      0.00       6.22     Page Write Rate       0.00       6.69      0.00      58.66     Page Write I/O Rate   0.00       0.27      0.00       1.66     Inswap Rate           0.00       0.00      0.00       0.00     Free List Size     3691.00    3604.09   3392.00    3771.00     Modified List Size  149.00      73.36      4.00     181.00 

No attempt is made to explain these displays, only to illustrate the number and variety of categories and presentation methods. The Compaq/HP documents explain these categories very well.

I have spent some time describing various algorithms and data structures used by OpenVMS in my book OpenVMS Operating System Concepts. The point is that from these displays (and others like them), an observer can focus on the resources that are overtaxed. Once the problem resource is identified, various solutions can be proposed, implemented, tested, and compared.