LAHRPreC: A Novel LSTM-ARIMA based Hybrid Resource Prediction Model for Cloud

Abstract

Dynamic resource planning is a critical task to ensure the Quality of Services (QoS) in a cloud environment. To optimally utilize cloud resources, it is required to accurately predict future demands in a real-time environment. In the cloud computing paradigm, resources like servers, networks, and cloud storage can be allocated to end-users dynamically based on their demand. Since the cloud workload is massive as well as heterogeneous in terms of various attributes and cloud resource demands fluctuate, cloud service providers (CSP) are required to efficiently furnish the available resources. In this paper, we proposed a hybrid approach named LAHRPreC to predict future resource demands in cloud environments and which will affect the resource administration. LAHRPreC employs long short-term memory (LSTM), a deep learning model, and auto-regressive integrated moving average (ARIMA), a statistical analysis model. The proposed model is less complex and significantly improves accuracy and execution time.