Chinasoft has been using SAS system in various applications since its foundation. The platforms involved range from pc-server, minicomputer to IBM mainframe. Some senior technical staff have been using SAS for 12+ years to develop sophisticated SAS applications. Besides the deep understanding of the execution mechanism of the mainframe SAS, we also have adequate command of knowledge about mainframe operating system, such as TSO, CLIST, ISPF, JCL, PDS and VSAM, etc.

A sample case where mainframe involved is the ERP (code name as 9672Project) system at one leading steal group in China. Its architecture is shown in the following figure.

The production system for manufacturing and sales are deployed on the IBM mainframe (IBM 9672). It is the main data source of the ERP system. The SAS applications deployed on the mainframe implement the following tasks:

1. Access data from DB2 periodically and /or upon administrator’s instruction;
2. Data cleaning and integration;
3. First-stage data processing, such as lightly summarization;
4. Generate and distribute the pre-designed regular reports;
5. Function as the data provider to the Baosteel enterprise data warehouse deployed on the very high grade Unix server.

On the mainframe, we use TSO commands and CLIST to develop, organize and invoke interactive applications; ISPF facilities are used as part of development environment; JCL and SDSF are used to define, control and monitor the execution of streamline or concurrent batch data processing and to backup jobs; System utilities (IEBGENER, IEBCOPY, IEBUPDATE, IEHLIST, etc.) and IDCAMS are used to manage and process PDS and VSAM data sets.

Once data (approximately 1G of new data generated per day, several tegabytes in total) entering the data warehouse, many other data analysis applications can be accessed by users ranging from production line operators, quality inspectors, salesmen, to executives. Through technologies such as reports and graphics, OLAP, statistics and forecasting, data mining, these applications give great supports to manufacturing management, product management, CRM/SRM, sales and marketing and quality control, etc.

Comprehensive SAS functions are involved in the development, such as data steps to handle file input and output, implement sophisticated data processing logic; and lots of SAS procedures are used to handle data integration, aggregation, sorting, merge, transformation, partitioning, data analysis and macro facilities, SCL is used to organize execution logic, auto-generation of codes; access interface (libname, procedures and pass-through facilities ) to handle data extraction from relational database.

The subjects of analysis including:

1. Aided-Design for New Products
   Building statistical models to detect and study the relationship between the capability differences of steel products with their major and micro chemical elements and manufacturing techniques. These analyses greatly shorten new product design circle and cut down investments on physical experiments.
2. Manufacturing Plan Forecasting and Optimization
   Designing statistical and time-series models to forecast production amount and demands on raw materials to make manufacturing plan more reasonable and manageable.
   Moreover, we use mathematical planning algorithm to optimize the manufacturing plan, for example, to find out best way of transporting raw-materials to processing plants and mid-way storage of half-finished products.
3. Pre-warning for Dumping
   Collecting data on types of iron-steel products, prices, sales amounts, manufacturer from the market to build antidumping warning database. Building industry-affection model to analyze the relationship between import quantity and price change with domestic production/market situation. Providing pre-warning information about potential import-dumping activities.
4. Supplier Management Analysis
   Analyzing the quality of products and services from the suppliers, combined with other information, such as price to score the suppliers, so that the predictive results can be referenced in future purchase orders.
5. Sales and Marketing Analysis
   Based on the data of manufacturing cost, productivity, market prices, quantities of inventory and historical profit, providing the analysis about the market strategies to improve overall sales benefit.
6. Customer Contribution Analysis
   Based on the data of historical orders, product cost, design, from various perspectives (profit, market percentage, reputation improvement) and the KPIs of customer contribution value, modeling the customer overall contribution. By using these models,we performed customer classification and VIP customers identification analysis.