Foundry Cuts User Response Time 75 Percent by Monitoring Mainframe Transactions with ASG-TMON™ for CICS/VSE®
Industry: Foundry
Location: U.S and UK
Grede Foundries, Inc., headquartered in Milwaukee, WI, operates 11 foundries in the U.S. and U.K. It is a recognized producer of high quality castings in gray iron, ductile iron, and steel for the automotive, agriculture, and construction industries.
The Challenge
Some of the electronic documents that Grede works with include invoices,
material releases, price catalogs, purchase orders, advance ship notices,
purchase order changes, material receipts, ship schedules, and text memos. The
company also exchanges design data with customers using CAD systems. It offers
current job status reports, online cost estimate requisitions, and current metal
pricing in the customer service section of its Web site. The backbone of Grede’s
data processing operations is a series of custom and packaged applications that
run on an S/390 IBM mainframe in the VSE/ESA environment. These applications are
responsible for managing all the information needed to fulfill Grede’s core
business objectives, such as meeting delivery promises to customers while
maintaining low inventory levels.
A few years back, however, users were experiencing delays in accessing these mission-critical applications. It would typically take about a second for a user to be able to gain access to the task queue. This was not a major delay but Rich Smrcina, manager of operations for Grede, recognized that when you multiplied this figure by the many times each of the company’s hundreds of users accessed the systems each day, the delay really added up. Standard CICS diagnostic tools gave him little insight into the problem.
"CICS has a built in facility called CICS Execution Diagnostic Facility (CEDF), but it is used primarily as a unit test debugging tool," Smrcina said. "I felt that I needed a full function monitoring tool to delve deeply into what I already suspected were rule-breaking programs eating up way more than their share of resources.”
The Solution
Smrcina selected ASG-TMON™ for CICS/VSE®
because it provided a comprehensive approach
to modeling a full range of mainframe application issues within a single, very intuitive
user interface.
“When I started using this tool, I found quite a few surprises,” Smrcina said. “First, I discovered that some applications were reading excessive amounts of information out of the database in a single request. Essentially, CICS was being used as a query tool. To address this problem, I capped the amount of data that could be addressed in a single transaction. Our programmers and users quickly got the message to be a little more considerate of system resources and before long nearly everyone was operating according to the new rules. Response time began to improve almost as soon as the monitoring system was implemented.”
Smrcina said that further improvements came from his ability to quickly identify and fix runaway transactions. “Normally transactions execute quickly enough that they do not even appear on the monitor,” he said. “Runaway transactions, on the other hand, typically stay up much longer, hurting performance or even bringing the system to a halt. TMON for CICS/VSE helps me investigate first by indicating the total amount of CPU time used by the transaction and which databases have been hit. This level of detail usually gives me a pretty good handle on what type of problem I am facing.”
Smrcina usually abends the program and reports the problem to the programmer that is responsible for the application. “TMON for CICS/VSE has a facility called Supertrace that helps me pinpoint the cause of a problem quickly and efficiently,” Smrcina said. “It can track down the path through the program to the abend and zero in on exactly which call was being executed when the abend occurred. I can easily correlate that call to the line in the code where the call was generated.”
Measurable Results
Smrcina also uses the same feature in the test environment, such as when a program
does not handle a condition properly and the programmer does not know where in the code
the condition is located. “I run Supertrace to search through the code to the point
where the program abended. It shows the program name that was running and the code path
through the program by the CICS calls. While CEDF would take minutes of clock time to
run this kind of trace, Supertrace only takes a few seconds. This saves time and
indirectly improves response time.”
The monitoring tool is also used to identify problems with VSAM LSR buffer pools. It clearly shows the degree to which buffers are being used, making it easy to determine whether they are over- or under-allocated. Smrcina said that the information provided goes far beyond what is available in CICS shutdown statistics. “The shutdown statistics don’t tell the whole story,” he said. “They can tell you if you are under-allocated, but TMON for CICS/VSE also tells you if you are over-allocated. The software also helps us manage storage resources.”
"We have improved the response time of our online operations while keeping our batch processing window within acceptable boundaries. The gains we made up front demonstrated the importance of comprehensive monitoring. As a result, we put a program in place that allows us to continually track and improve application performance." Smrcina concluded.
