Cloud based services have grown substantially due to the cost-effective migration of applications to the cloud. As a result, there are now a plethora of data centers that can provide these services on a massive scale, with a wide variety of the user experience and very little downtime. The need to control the power consumption and performance of the data center's constituent nodes without affecting service level agreements (SLAs) arises from the promise to provide differentiated services on a large scale. The efficiency of the power consumption in such data centers poses a significant challenge to cloud computing. Data center server energy consumption may be reduced by the use of various optimization methods, such as workload consolidation and machine location. Here, we provide a data-driven predictive neural network architecture that, at any given time in the future, can accurately predict the server's power consumption by taking into account all of its components in addition to the load of incoming requests.
Research Journals