A Dynamic, Novel Unified Methodology for predicting Cloud Workload on Cloud Computing

Abstract

In cloud, good resource management is critical, and carrying out the prediction of workload is a crucial step toward accomplishing that goal. As we know that the workload of task which is being processed for a long time can be predicted by the recurrences of their past workload, whereas it is much complex job to predict the workload of the task which doesn’t have a recurring pattern of workload. So, in this concept we are focusing on a method for workload prediction so the user/organization know when the workload is high or vice-versa. We have different approaches for workload prediction as in this,we use information about the workloads of a pool of tasks rather than the ancient workload for a task to predict the potential workload of that task, here we carry the knowledge about the workloads for a series of tasks to aid in the forecasting of new task workloads. As here we deal with the core of cloud computing model this is designed for carrying out the results in more efficient manner without any error in the process. It demands different factors for on and off premise of infrastructure for handling internet based applications. So in order to realize this concept, we have developed a clustering and an approach which is learning based. As we begin, the tasks are divided into several clusters. The approach Long short term which is the architecture of neural network will then be implemented to acknowledge the characteristics of every cluster’s workload. Hence, the results are predicted by the core of the technology which is much faster and efficient with minimal errors. Workload prediction also depends on the popularity of an application’s data.