Hybrid Round-Robin - Classification Algorithms for Cost and Time Optimization in Cloud Environment

Purpose: The classification algorithms (CA), including Decision-Tree (DT), K-Nearest-Neighbor (KNN), and Support-Vector-Machine (SVM) under the Machine Learning (ML) category, are integrated with the existing Round Robin (RR) algorithm to develop hybrid RR-DT, RR-KNN, and RR-SVM to enhance cost and time metrics in the cloud environment. Performance Metrics: To compare the proposed RR-CA algorithms with baseline$\mathbf{R R}$, cost, and time are utilized, along with measuring the improvement percentage (IMPR%) of RR-CA with RR results. Dataset: The cluster tasks of Google were utilized to compute using the proposed and RR algorithms. Experiment: Experiments were conducted across five Virtual Machine (VM) scenarios, ranging from fifty to one thousand, using the proposed hybrid RR-CA algorithms with the RR algorithm. Results: Considering the cost, the RR-SVM provided an IMPR% of 0.49% when compared to the RR algorithm. Other algorithms, RR-DT and RR-KNN, provided negative cost results, i.e.,$-0.15 \%$and$-0.06 \%$compared to the RR. On the other hand, the RR-DT provided an IMPR% of 0.02% in terms of time when compared to the RR algorithm. The RR-KNN and RR-SVM provided negative time results, i.e.,$-0.15 \%$and$\mathbf{0. 2 2 \%}$compared to the RR algorithm. Stability Validation: The obtained results were validated by conducting stability evaluation using five regression models. Findings: Enhancing the existing RR with SVM and DT provided improvements concerning cost and time, respectively. Future Work: The RR-SVM and RR-DT shall be integrated together with convolutional neural networks to further optimize the cloud and provide optimal results.