Installation Considerations - mfx - 3.1

The list below includes MFX facilities you should be aware of before beginning the installation process.   

Details on these items are in the sections below and more details are in Installation.

System Resident or Non-Resident Configuration

Either a system resident or non-resident release of MFX can be installed. When the resident release is installed, multiple sorts can share the same MFX load modules. Performance can be improved, and overhead can be reduced.   

All the resident modules of MFX will be installed above the 16-megabyte line. In this way, more space is available in the Link Pack Area for programs which must reside below the 16-megabyte line.   

The BetterGener Facility     

BetterGener is a high-performance, transparent copy facility, which, in many cases, can replace IEBGENER. When BetterGener is activated, eligible jobs are automatically processed by MFX. You do not have to change IEBGENER job streams in order to take advantage of MFX’s much more efficient copying techniques. BetterGener will allow your installation to achieve impressive reductions in CPU time and EXCPs.   

PARASORT and Special Esoteric Unit Names    

PARASORT improves elapsed time performance for sorts whose input is a multi-volume tape data set and/or concatenated tape data sets. Reduced elapsed time can help critical sort applications achieve batch window goals.   

The performance improvement from PARASORT is a result of processing the SORTIN input volumes in a parallel fashion. Depending upon the resources provided, elapsed time can be reduced up to 20% for 2-way input and up to 33% for 4-way input.   

PARASORT requires additional tape units for the application. You will need from two to eight times the current number of tape units, depending upon resource availability and the degree of improvement desired. PARASORT automatically manages the tape units and minimizes the use of the tape drive resources by deallocating excess tape drives during initialization and releasing all the extra units at the end of the sort input phase.   

The additional tape units are defined to PARASORT on up to four DDs labeled SORTPAR1, SORTPAR2, SORTPAR3, and SORTPAR4. A segment of SORTIN will be read in parallel from each of these DDs. The segmentation of SORTIN is automatic.   

Increased parallel input processing (up to four SORTPARn DDs) increases the elapsed time benefit. However, for optimal PARASORT performance, MFX must be able to read each SORTPARn input DD simultaneously with no channel contention.   

Generally, the normal allocation of tape drives will not ensure sufficient channel path availability, particularly for a 4-way PARASORT (four SORTPARn DD statements). Therefore, you may need to create special esoteric unit names for PARASORT. Assigning certain groups of drives to each esoteric unit name used in the PARASORT JCL will in most cases ensure the required channel separation.   

Although you may need to create special esoteric unit names, PARASORT itself is included with MFX and will be available once MFX is installed. If you will be using PARASORT, see the instructions for creating PARASORT esoteric unit names in PARASORT.

Dynamic Storage Management (DSM)             

MFX’s Dynamic Storage Management (DSM) capability is an automatic facility for the dynamic control of memory utilization and SORTWK device selection.   

Memory resources can come from the address space, data space and memory objects (which are used by MFX’s ZSPACE technique). The DSM facility dynamically determines which resources to use and how much of them to commit to a sorting application. DSM is designed to provide the best performance for each sort while optimizing overall system throughput.   

The DSM facility considers the VSCORET option to be an initial recommendation for the amount of address space memory to use. From this starting point, DSM’s sophisticated algorithms analyze overall system activity and the particular sort’s resource requirements. Using this information, the actual amount of address space, data space and memory objects is determined at run time.   

For example, during periods of low activity on the system, the DSM facility makes efficient use of idle memory, providing better performance for the particular sort that’s running and improving throughput for the entire system. On a busy system (as when an interactive facility is running), MFX utilizes less memory so that more remains available for other system tasks.   

In the area of dynamic SORTWK device selection, DSM chooses devices from among those designated for its use on the basis of speed and the level of contention for those devices. DSM looks at contention from all applications and from other sorts currently running on the system.   

The DSM facility can be operated in one of two ways. The most effective way is through a centralized administrative program that executes in its own address space and communicates with sorts in progress through the z/OS subsystem interface. The term “global DSM” will be used to refer to this preferred way of using DSM throughout this manual.      

A second method of using DSM is locally from within a particular execution of the MFX product itself. This method will be referred to as “local DSM” and can be used in all operating environments. The resources coordinated by local DSM are a subset of those handled by global DSM, and its decisions about resource usage are based on local rather than global considerations. If global DSM is active, it will supersede the local DSM.   

Global DSM Functions        

Global DSM has two functions: monitoring and decision-making.

Monitoring

When global DSM is active, it continually monitors the state and performance of SORTWK devices, central storage, and DASD I/O channel paths. The knowledge it acquires through monitoring is recorded in a special history database, allocated as a single small data set. The information in the database summarizes the patterns and regularities that govern the daily and weekly cycles of system activity. Since monitoring is an ongoing process, the database will always contain an accurate and up-to-date profile of the system.

Decision-Making

Global DSM’s second function is to decide how each sort can make the best possible use of available resources. Global DSM’s algorithms analyze and coordinate the information in the history database, the current level of resource usage, and the sort’s own characteristics. In this way, global DSM selects the best way to balance available system resources and run each sort as efficiently as possible while improving overall system throughput.   

Although global DSM considers virtual storage in the sort’s address space, data space, memory objects, and SORTWK, its true focus is on the underlying real resources: central storage and available SORTWK disk devices and their associated channel paths.            

Local DSM   

Local DSM is a variation of MFX’s DSM facility in which the information used for DSM’s decision making is limited primarily to the information available in the sort’s address space. This includes the characteristics of the specific sort and the current availability of system resources, but does not include the system monitoring and history information used by global DSM. While local DSM has the ability to perform some basic system-wide coordination of storage resources between MFX applications, its scope is limited compared to global DSM. The decisions made by local DSM will improve individual sort efficiency and system throughput, but not as effectively as the decisions made by global DSM.   

Local DSM is controlled by parameters of the DSM and DSMWEND options in SYNCMAC.