Browse Prior Art Database

High performance design of MRP engine

IP.com Disclosure Number: IPCOM000129121D
Original Publication Date: 2005-Sep-28
Included in the Prior Art Database: 2005-Sep-28
Document File: 3 page(s) / 33K

Publishing Venue

IBM

Abstract

Disclosed is a software design for realizing high performance of Material Requirements Planning (MRP for short). In contrast with conventional array table oriented design, it is based on a linked list oriented design, reading all data necessary to execute MRP calculation into memory. While MRP systems take many hours for completing their execution in that they must handle huge volume of data, this new design reduces the execution time considerably from hours to several minutes.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 47% of the total text.

Page 1 of 3

High performance design of MRP engine

Disclosed is a software design for realizing high performance of Material Requirements Planning (MRP for short). In contrast with conventional array table oriented design, it is based on a linked list oriented design, reading all data necessary to execute MRP calculation into memory. While MRP systems take many hours for completing their execution in that they must handle huge volume of data, this new design reduces the execution time considerably from hours to several minutes.

Followings are the outline of the flow of the new design of MRP.

1. Read all input data of MRP from database into memory area with minimum query conditions.
2. Register each item data into a hash table, and implement the parent-child relations among items in terms of an adjacency list data structure, which is a variant of graph data structures. The adjacency list is, for the sake of efficiency, combined with a hash data structure, constructing a hashed adjacency list data structure.
3. Link each requirement data and order data to their proper item data in the form of linked lists.
4. Sort every linked list by means of merge sort. In accordance with the characteristics of input data, merge sort can be replaced by insert sort.
5. Arrange the linked list of requirements data so as to resuce the number of list nodes to be traced as their calculation proceeds.
6. Traverse the graph of the hashed adjacency lists for B/M explosion without any search of necessary data.
7. Prepare phantom stack for explosion of phantom items for its efficiency.

Structure of P/N Data

 Since the unit of MRP calculation is performed for every item data, this is read first of all and set as axis data, around which all the related data of gross requirements, order, including product structures are collected in the form of nodes of linked lists whose heads are pointed by their axle item data node. Suppose, for example, that item A has two product structures of items B and C, two gross requirements, and two orders, they are deployed in memory in the data structure of linked lists as illustrated in fig. 1.

Fig.1

Hash Search of Item Data

 In order to construct the data structure illustrated in fig.1, it is necessary to search for the axis item data. The new design employs hash search of direct chain method. In production control systems, since parts number (P/N for short) of every item is treated as character strings, it becomes in case difficult to construct effective hash function that generates variously spread hash values, if P/N strings consist of very limited kinds of characters. As in ordinary production control systems, item data are, without any ordering conditions, apt to be retrieved with sorted order of P/N string, it is possible to construct perfectly balanced binary tree whose height can be limited to O (N log2N ), where N stands for the number of all data.

Fig.2

1

[This page contains 2 pictures or other non-text objects]

Page 2 of 3

 The new design...