S. Ananthi and P. Vaishnavi. Resource Provisioning for Reduced Reallocation on High-Dimensional and Highly Variable Cloud Workloads. Dynamic Systems and Applications 30 (2021) No.8, 1318 – 1334

https://doi.org/10.46719/dsa20213087

ABSTRACT.
Resource provisioning in cloud computing is a major component that can improve the performance of a cloud system to a huge extent. High dimensionality and high variability in the cloud workloads pose major challenges in the allocation process. This work presents an architecture that performs resource provisioning based on demand prediction and range-based resource allocation that ensures reduced reallocation. The proposed architecture is composed of two modules: the initial module contains a Tree based Ensemble Cascade (TEC) for effective demand prediction, and the second module contains a Modified Firefly Algorithm (MFA). MFA is used for range-based allocation of virtual machines. It ensures that the virtual machines falling within the requirement range is selected for usage. Simulations were performed with Cloud Sim using the Alibaba and the Google cloud traces. Experiments indicate effective workload predictions with reduced reallocations. Comparisons were performed with LPAW model from literature and results show a significant reduction in MSE values between 29% and 98% on the Alibaba cloud trace and between 92% and 98% on the Google cloud trace. The effective reductions in MSE levels depict accurate resource allocations, hence the model is considered to effectively reduce reallocations.

Keywords: Resource Provisioning, Cloud Computing, Ensemble Model, Firefly Algorithm, Workload Prediction.