Parallel Processing
Parallel processing is a method that distributes work that an application performs across multiple processors within a CPU. The parallel processing method achieves significant performance gains that are limited only by the number of processors that are available on the server. Thus, parallel processing reduces the time that a system requires to complete the work. For example, if you have excessive data volume, such as large bills of material, parallel processing can reduce processing time. These conditions determine whether the system would benefit from parallel processing:
Server with multiple processors.
Large volume of data at any level of MRP.
A method of parallel processing is to create an application that divides data into separate, independent paths. The paths process data simultaneously on multiple processors. The benefit of this method is that the application automatically divides data among processors and also eliminates data administration. The MRP/MPS Requirements Planning program and the Master Planning Schedule - Multiple Plant program use this method of parallel processing.
When you run the parallel processing versions of the MRP/MPS Requirements Planning and Master Planning Schedule - Multiple Plant programs, you plan items concurrently within the same low-level code because the items do not depend on each other. The system processes items with a low-level code that equals 1. The system processes all of the items at the current level before it can process any items at the next level: current level + 1. The system repeats this process until it has processed all of the items. Also, if a server has multiple processors, you can plan multiple items concurrently with the same low-level codes.
Parallel processing uses the software's subsystem capabilities to run multiple copies of a batch program. The system provides a unique subsystem version of a Universal Batch Engine (UBE) that enables you to specify the number of available processors on the server. When the system runs the batch program, preprocessing fills the subsystem table and begins the subsystem jobs. The jobs run the same version of the application and access the same processing option values. Each record in the subsystem table contains data that the system passes to a report. Each subsystem job processes records from the same table, one at a time. After the system processes all of the records, the system terminates the subsystem jobs.
The system has several applications that parallel processing supports. To use the parallel processing versions, you must perform these tasks:
Indicate the maximum number of the same subsystem jobs that the system can process at the same time.
Add or modify the value, UBESubsystemLimit=N, in the UBE section of the jde.ini file.
N represents the maximum number of similar subsystem jobs.
Set up multithreaded job queues.
Note: Technical support must complete standard tasks in system setup before you can perform the parallel processing method.
The MRP/MPS Requirements Planning and Master Planning Schedule - Multiple Plant programs have a processing option, Number of Subsystem Jobs, on the Parallel tab that indicates how many concurrent jobs can process the MRP data. You must use subsystem parallel versions that are specifically designed for parallel processing.