Recalibration of Analytics Workflows

As business decisions and strategies become more and more automated, real-time, and data-driven, enterprises need to create, manage and execute end-to-end analytics workflows that process increasing data volumes, from new heterogeneous data sources, on specialized processing engines. Workflows become more complex and time-consuming to design and execute, since they span a variety of systems and the amount of data being processed grows. Therefore, it becomes increasingly difficult to debug workflows in order to handle errors, as well as adjust the workflow design and calibrate task parameters for applications that perform exploratory data analysis. Towards this end, the workflowmanagement system should provide recalibrationmethods i.e. methods to monitor and to influence workflow processing at runtime. We demonstrate novel manual and automatic recalibration techniques for analytics workflows, on real use cases and data from the telecommunication domain and web analytics, but also on synthetic use cases and data.