Forecasting capacity available for processing workloads in a networked computing environment

Abstract

Embodiments of the present invention provide an approach for forecasting a capacity available for processing a workload in a networked computing environment (e.g., a cloud computing environment). Specifically, aspects of the present invention provide service availability for cloud subscribers by forecasting the capacity available for running or scheduled applications in a networked computing environment. In one embodiment, capacity data may be collected and analyzed in real-time from a set of cloud service providers and/or peer cloud-based systems. In order to further increase forecast accuracy, historical data and forecast output may be post-processed. Data may be post-processed in a substantially continuous manner so as to assess the accuracy of previous forecasts. By factoring in actual capacity data collected after a forecast, and taking into account applications requirements as well as other factors, substantially continuous calibration of the algorithm can occur so as to improve the accuracy of future forecasts and enable functioning in a self-learning (e.g., heuristic) mode.