In this pattern, learn how to analyze batch jobs on a mainframe with IBM® Watson™ Machine Learning on z/OS®. This includes ingesting batch jobs’ operation data from SMF Type 30 data, exploring periodicity and correlation, and predicting elapsed time and detecting anomalies.
Running batch jobs is a critical operation on a mainframe. Every day there are approximately 10000 – 60000 jobs running. At the first or last day of a month, quarter, or year, the workload might double that. Operating a mainframe in an optimal status with high productivity and the ability to handle business demands is required.
Machine learning analytics on large volume logs of batch jobs might help you in four areas:
- Understanding batch job seasonality and workload change trends
- Gaining insight of impacts for batch jobs
- Predicting the elapsed time of long batch jobs
- Identifying abnormal candidates in batch job instances
- You could work on Watson Machine Learning for z/OS through a web browser.
- Watson Machine Learning for z/OS provides a Jupyter Notebook for you to code in Python, Scala, and R.
- Watson Machine Learning for z/OS provides a modeler flow for you to explore data and train models.
- You could read z native files, for example, an SMF Type 30 file with a Python notebook based on the included mainframe data service of Watson Machine Learning for z/OS.
Find the detailed steps for this pattern in the README file. The steps show you how to:
- Download project content as a .zip file (for Windows™ or Mac) or .tar.gz file (for Linux®).
- Log in to Watson Machine Learning on z/OS.
- Create a new project.
- Add the downloaded .zip file.
- Check the project content.