Using Machine Learning to Reduce Risk for Workload Automation
September 20, 2018 2:30 pm
Traditional analytics (such as Google Analytics) are useful for understanding historical activity and performance. Unfortunately, this type of analytics does not provide a clear picture of what is likely to happen in the future. In today’s modern enterprise, this foresight is crucial for planning and operating a workload automation environment. This is where traditional analytics fall short.
Advanced Predictive Analytics
Terma has been using machine learning to provide predictive analytics for workload automation for over a decade. All of the predicted data generated by The TermaAnalytics platform is continuously informed by new information as it arrives in near-real-time.
Machine learning is used throughout Terma products. Some important examples are:
- Duration (how long a job or task will take)
- Latency (lag between scheduled time and actual start time)
- Unscheduled events (things that run, but have no defined reason to do so)
Machine Learning to Predict Duration
While modern systems are extremely fast, no process runs in zero time. Terma applies increasingly sophisticated analysis to historical durations for a job or task to make an accurate prediction of the future. Terma’s recently-released “adaptive” predictions automatically apply multiple statistical models against historical data to choose the most accurate model when predicting the next run for every job in the system.
Machine Learning to Predict Latency
All systems have latency. In large systems, latency can be a significant factor in meeting critical service-level agreements or SLAs. Not only are latencies sometimes very large, often they appear sporadic or random. Terma captures and analyzes historical latency to provide far more accurate predictions than are possible by duration alone.
Machine Learning to Predict Unscheduled Events
The nature of large systems is that they increase in complexity over time. As systems become more complex and interconnected, events occurring outside of the scheduling system have a material effect on critical workflows. This could be the transmission of files from vendors, processes from other systems initiating a process inside a scheduler, or any number of other examples. Terma has developed a set of features to detect these external events, manage, and predict them accurately. Using classic machine learning techniques, Terma makes recommendations, and can automatically apply and adjust predictions for these seemingly random events.
Bringing it all Together
Terma’s ability to predict unscheduled events, combined with accurate details about latency and duration, provides unprecedented accuracy in predicting workload environments. A predictable environment with end-to-end visibility enables accurate forecasting of SLA breaches, root cause, and impact analysis, before, during and after an adverse event. Know your risk, the impact to the business, and potential remedies in real-time with the TermaAnalytics platform.
To learn how Terma’s predictive analytic solutions can help you optimize your workload environment, contact Terma Software for more information.