This paper focuses on batch environment of an enterprise. We use various data sources such as dependency data, alert and ticket data, and run history data of the jobs in batch systems to build a model of batch operations.
We enrich the model with various attributes derived by graph analysis of dependency data, time-series analysis of execution history of jobs, and text and statistical analysis of alerts data. We present reactive and proactive approaches for improving stability of batch operations using this model and demonstrate their effectiveness through a real-world case-study.