Improving Service Broker Policy of the Cloud Using Reinforcement Learning Through Equally Spread Current Execution Load Balancing Policy

The cloud in the modern era has been one of the central pillars of Information Technology by providing pragmatic solutions. The Service Broker Policy (S_B_P) implemented along with the Load Balancing (L_B) strategies used in the cloud play a fundamental role in providing these solutions consistently. Hence, evaluating the performance of these S_B_Ps used by the L_B strategies becomes significantly essential. The primary aim of this paper includes an experimental study to examine the performance of the S_B_Ps of the cloud, namely Closest Data Centre (Cl_D_C), Optimize Response Time (Op_R_T), and Reconfigure Dynamically with Load (Re_D_w_L) using the Equally Spread Current Execution Load Balancing (E_S_C_E_L_B) strategy. To do so, an experiment was conducted in a cloud simulator where jobs were computed in the cloud DCs in several scenarios. The performance parameters considered for this performance evaluation of SBPs are Overall Response Time (Ov_R_T) and Data Centre Processing Time (D_C_P_T), both measured in milliseconds (ms). The results obtained from this experiment convey that the S_B_Ps Cl_D_C, Op_R_T, and Re_D_w_L take an average Ov_R_T of 1324.69 ms, 1324.47 ms, and 4609.29 ms, respectively. Concerning D_C_P_T, the S_B_Ps Cl_D_C, Op_R_T, and Re_D_w_L output an average of 1024.61 ms, 1024.33 ms, and 4324.34 ms, respectively. Hence, the performance of the S_B_P Op_R_T outperforms the other S_B_Ps Cl_D_C and Re_D_w_L, concerning both the performance parameters Ov_R_T and D_C_P_T. Lastly, a model of Reinforcement Learning – S_B_P Op_R_T has been presented to provide intelligence and improve S_B_P of the cloud.