Making Batch Systems Predictable

In today's IT world, batch workload is responsible for all the back-end operations. Batch systems are defined by a sequence of batch jobs. These jobs are usually non-interactive in nature and their execution is defined by precedence relationship. To ensure smooth and stable operations, it is important to predict a future batch. Batch management is inherently reactive in nature as it lacks end-to-end understanding leading to an increasing reliance on intiution. It is important to reimagine batch operations by bringing a paradigm-shift from today's intuition-driven reactive approach to analytics-driven proactive approach. This can be made possible by fine-grained modeling of batch systems, an intelligent layer of algorithms, and augmenting analytics with knowledge of the domain of the batch.