SCA (Speed Catalog Access) is a software product for speeding up catalog management of SF pubsets. SCA replaces the sequential search of catalog entries by means of direct access. SCA does not change the catalog structure.
This procedure is implemented with tables:
The freelist table provides information on free space in the catalog blocks. The freelist table is managed by routines which run under the calling task.
The cross-reference table assigns the logical block number of the catalog block containing the entry to the file/job variable names. Irrespective of the variant set (see "SCA execution variants)"), the cross-reference table is managed either by the SCA task or by routines which run under the calling task.
The SCA tasks of different catalogs can be simultaneously started and terminated.
Using SCA enables the total system throughput to be increased and greater flexibility in data center organization:
The runtime for programs which use the catalog intensively (e.g. ARCHIVE) and the response times of commands with a high percentage of catalog accesses (e.g. CREATE-FILE, MODIFY-FILE-ATTRIBUTES, IMPORT-FILE) can be considerably improved.
It is no longer necessary to catalog files with frequent catalog accesses at the front of a user chain.
The following should be noted:
Noticeable performance improvements can only be expected if the users make heavy use of the catalogs (> 60-100 file entries per user ID).
SCA requires additional resources at runtime, such as virtual and real memory, tasks, etc.
SCA use can be reduced to a few pubsets if the file catalog structures are optimum.
Command | Meaning |
ADD-MASTER-CATALOG-ENTRY | Create an entry in the MRSCAT catalog list on the home pubset with definition of whether SCA is to be started automatically when IMPORT-PUBSET is called |
ENTER-JOB | Start and terminate SCA |
MODIFY-MASTER-CATALOG-ENTRY | Modify an entry in the MRSCAT catalog list on the home pubset with definition of whether SCA is to be started automatically when IMPORT-PUBSET is called |
Table 24: SCA command overview
SCA is installed using the installation monitor IMON and requires the following files under the TSOS ID
SYSLNK.SCA.<ver> | SCA module library for SUs /390 |
SYSREP.SCA.<ver> | REP file |
SPEEDCAT.ENTER.START | Start procedure |
SPEEDCAT.ENTER.STOP | Termination procedure |
SPEEDCAT.START | Startup program |
SPEEDCAT.STOP | Termination program |
SCA execution variants
When SCA is started for a specific pubset, the user can specify which SPEEDCAT variant is to be loaded:
SCA with task swapping
A separate SCA task is set up when SCA is started. It generates the SCA tables and waits for jobs. The SCA task communicates with the user task via a double-chamber bourse, in order to exchange requests and results.
Allowance must be made for performance degradation caused by task swapping with this variant. The load on class 4 memory is, however, reduced since the SCA directories are held in class 5 memory (where they are only read by the SCA task).SCA without task swapping
All SCA tasks can be executed without task swapping. A separate SCA task is also set up in this case when SCA is started, but this is terminated once the tables have been built up.
The cross-reference table must be accessible here to all user tasks and is created in class 4 memory for this reason. Access is regulated with exclusive locks via a signal bourse (single-chamber bourse) to prevent several tasks from simultaneously reading from and writing to the directories.
Starting SCA implicitly with IMPORT-PUBSET
If SCA is to be executable during IMPORT processing, this must be stored in the pubset's MRSCAT entry. Systems support can create an appropriate entry with the START-SPEEDCAT operand in the ADD-MASTER-CATALOG-ENTRY or MODIFY-MASTER-CATALOG-ENTRY command. The default value for starting SCA in the ADD-MASTER-CATALOG-ENTRY command is START-SPEEDCAT=*AUTOMATIC, i.e. SCA is to be started automatically.
Starting SCA explicitly during the session
Start with catalog ID specification:
ENTER-JOB FROM-FILE=SPEEDCAT.ENTER.START,JOB-NAME=SCA<catid> where <catid> is the single-digit or multiple-digit pubset catalog ID. The SCA variant is set during response to message
DMS03FE.Start without catalog ID specification:
ENTER-JOB FROM-FILE=SPEEDCAT.ENTER.START where input of the desired pubset catalog ID is prompted for on the console. The SCA variant is set by the additional response to message
DMS03FE.
It should be noted that SCA should only be started for a pubset after the pubset is completely imported. The command for starting SCA during the session can also be integrated into the CMDFILE.
Terminating SCA
Termination with catalog ID specification:
ENTER-JOB FROM-FILE=SPEEDCAT.ENTER.STOP,JOB-NAME=SCA<catid> where <catid> is the single-digit or multiple-digit pubset catalog ID.
Termination without catalog ID specification:
ENTER-JOB FROM-FILE=SPEEDCAT.ENTER.STOP where the pubset catalog ID is prompted for on the console.
Implicit termination:
The EXPORT-PUBSET and SHUTDOWN commands terminate the SCA task or release the class 4 memory used for the cross-reference table and freelist.
The command for terminating SCA can be stored in a RUN command sequence.
If the catalog ID is not specified (2), the commands may also be issued by the operator.
Error handling
SCA deactivates itself automatically and normal catalog access is reverted to if an error occurs in SCA or if hardware-related write errors in the catalog occur. A system dump is initiated, depending on the type of error.
SCA uses the SERSLOG function (see the “Diagnostics Handbook” [14]) to output diagnostic information. Error messages are also output to the console.