The SYSTEM ID Parameter
The name you specify as the system-name on the SYSTEM ID parameter of the SYSTEM statement is the identifier of the DC/UCF system you are defining. System-name is the nodename of the DC/UCF system and should be specified as the object of a nodename variable as appropriate. Additionally, system-name is the name you should specify when creating the resource SYSTEM as part of CA IDMS central security set-up.
Using AUTOTUNING
To use auto-tuning, change tracking must be active for the DC/UCF system, since statistical information needed for tuning is captured in the SYSTRK files.
New DCMT commands display and reset values used in tuning.
While the use of auto-tuning can be beneficial in many situations, it is not recommended under the following conditions:
Choosing an Area Acquisition Threshold
Before the area acquisition threshold count is reached, the system will free locks on areas previously locked by the database transaction if the transaction must wait to place a lock on another area. Once the threshold is reached, the system will not release existing area locks before waiting for a new area lock.
You should specify the default of OFF RETRY FOREVER unless experience shows that a database transaction is not gaining access to areas as needed.
Choosing a Deadlock Detection Interval
Always set the DEADLOCK DETECTION INTERVAL parameter to 1.
System-Internal Parameters
DPE COUNT is a system-internal parameter. You should accept the default value unless experience indicates that this value is insufficient to meet processing needs.
ECB LIST is a system-internal parameter; you should accept the default value unless experience indicates that this value is insufficient to meet processing needs.
RCE COUNT is a system-internal parameter; you should accept the default value unless experience indicates that this value is insufficient to meet processing needs.
RLE COUNT is a system-internal parameter; you should accept the default value unless experience indicates that this value is insufficient to meet processing needs.
The ECB List
The ECB list is used to synchronize events between the DC/UCF system and the host operating system. The list contains pointers to ECBs for each task waiting for an operating system event (for example, a disk read). Upon completing an event for which a task is waiting, the operating system posts the appropriate ECB. The DC/UCF system can then dispatch the waiting task.
Journaling Retrieval Run Units
Specifying NOJOURNAL reduces the size of the journal file; however, potentially valuable audit-trail and statistical information is not recorded.
Regardless of the JOURNAL/NOJOURNAL specification, the system always writes the following records to the journal file:
Specifying a Journal Fragment Interval
If your journal files are large, a journal fragment interval can significantly enhance warmstart processing. The warmstart logic begins its recovery processing at the most recently accessed journal fragment. Additionally, used with the journal driver task, a journal fragment interval can result in journal records being written to the journal more quickly thus reducing journal file bottlenecks.
To activate the journal fragment interval, you must specify a journal fragment interval of at least 100.
Specifying a Journal Transaction Level
You can reduce journal I/O activity by activating the journal transaction level and specifying the point at which CA IDMS will defer the writing of journal buffers to the journal file. Instead, CA IDMS will wait until the journal buffer is full or until the number of active transactions drops below the journal transaction level, before writing the journal buffer to the journal file.
Specify at least a journal transaction level of 4.
Specifying a Print-Screen Key
The print-key specification overrides the control-key assignments in the keys table used for online products in the CA IDMS environment. For example, if the keys table for online IDD assigns PF12 the function of printing the entire work file and the SYSTEM statement names PF12 as the print-screen key, PF12 will print only the current screen during an online IDD session. For more information about keys tables, see KEYS Statement.
Note: The transfer control facility (TCF) uses PF3 and PF9 for the TERMINATE and SUSPEND functions. If you define the system-wide print-screen key as PF3 or PF9, be sure to specify an override for tasks that execute under TCF. You use the system generation TASK statement to override the print-screen key for individual tasks. For more information about the TASK statement, see TASK Statement.
Adjusting the TCE Stack Size
The recommended value for the STACKSIZE parameter is 1200. Use this value unless experience indicates it is insufficient to meet processing needs. Runtime task abend code D009 and system abend code 3973 indicate that the storage stack size requires adjustment.
No CA IDMS SVC under z/OS or z/VSE
If the SYSTEM statement specifies SVC IS NO under z/OS or z/VSE:
The System Trace Facility
The system trace facility is used to trace system events during the development and debugging of user programs and can be managed at run time with the DCMT DISPLAY SYSTRACE and DCMT VARY SYSTRACE commands. When the system trace facility is enabled, each database or teleprocessing request is traced through the system service modules used in processing the request.
System trace information is recorded in a trace table buffer allocated at system startup. Each entry in the trace table contains the task id, the request type, and the contents of registers 11 through 16 and 1 through 8 for a service request. When the trace table becomes full, the system begins recording new entries at the beginning of the table, overwriting previously written entries. When a program issues a SNAP request, the system writes the trace table to the log.
Automatic Program Definition
Null PDEs are used for the automatic definition of programs not defined by PROGRAM statements during system generation. When you execute a program that is eligible for automatic definition, the system uses a null PDE for the program definition.
Programs that are not eligible for automatic definition must be explicitly defined either during system generation by means of the PROGRAM statement or at runtime by means of the DCMT VARY DYNAMIC PROGRAM command.
More Information:
The User Trace Facility
The user trace facility is used for debugging purposes.
When the user trace facility is enabled for a terminal, information on requests issued from the terminal is recorded in the user trace buffer. When the buffer becomes full, new entries are written to the beginning of the buffer; previously recorded entries are overwritten.
The user trace facility is enabled at runtime by means of either the DCUF USERTRACE command or the DCMT VARY LTERM command.
Specifying commit and rollback options:
You can specify options that control the following commit and rollback behavior.
You can control whether a COMT or ENDJ journal record is written on a commit operation in which the database session remains active. Writing an ENDJ can reduce recovery time because less data has to be examined to locate the start of a recovery unit. This benefit applies to online recovery, warmstart, and ROLLBACK and ROLLFORWARD recovery operations. ENDJ is most beneficial in cases where long-running transactions perform relatively infrequent updates between commit operations. In cases where update transactions are committed frequently, writing ENDJ journal records can negatively impact the amount of information journaled because another BEGN journal record will have to be written before the first update following a commit operation.
Note: ENDJ journal records are always written when system rununits are committed, regardless of the ON COMMIT option specified.
You can control whether a new local transaction ID is assigned following a commit or rollback operation in which the database session remains active. Assigning a new transaction ID reduces the chance of duplicate IDs should this value wrap within a single cycle of a central version. It also has the effect of recording journal statistics separately by commit recovery unit rather than across all recovery units within a database transaction. You can assign a new ID on a commit operation only if you also specify that an ENDJ checkpoint record be written.
Note: A new transaction ID is always assigned when system run units are committed or rolled out.
Task Snap Dump and System Snap Dump Options
A task snap can provide useful information when developing and debugging user programs. These options can be overridden dynamically by the DCMT command at the Task and Program level.
|
Copyright © 2014 CA.
All rights reserved.
|
|