Characterisation of Hidden Periodicity in Large-Scale Cloud Datacentre Environments
John Panneerselvam, Lu Liu, Nick Antonopoulos · 2017
Given the energy consuming characteristics of the Cloud datacentres, reducing the associated energy implications of the datacentres has become a natural demand for the providers to promote eco-friendly datacentre execution. The characteristics of Cloud-based workloads are still not perfectly clear and the extensive level of heterogeneity found among both the Cloud workloads and the server resources impose various levels of intrinsic and extrinsic complexities in accurately characterising and modelling the workload behaviours in close correlation with the user activities. To this end, this paper conducts extensive analysis on the workload and user behaviours in a large-scale datacentre environment, with the motivation of exhibiting the periodic characteristics of Cloud Computing. Both the workloads and the users at the datacentre are analysed by subjecting their behaviours to various periodical effects. Important contributions of this research work is the characterisation of the workload and user behaviours at the datacentre in terms of the job arrival frequency, recurring trend of job submissions, session usage and job diversity. The inferences presented in this paper are believed to provide sufficient knowledge to the Cloud providers to achieve optimised scaling and allocation of server resources for promoting sustainable and eco-friendly datacentre execution in accordance with the workload and user behavioural trends.