iRD-NN: an improved RainDrop-driven Neural Network model for cloud workload prediction

Abstract The rapidly evolving landscape of the cloud computing necessitates development of an accurate workload prediction mechanism to ensure an efficient resource utilization, flexibility and minimal operational cost. However, the unforeseeable workload patterns in terms of sudden spikes in the upcoming requests, redundancy and prolixity in clients’ demands, often impedes the workload prediction accuracy. This paper introduces a novel and improved RainDrop-driven Neural Network ( i RD-NN) workload prediction framework leveraging the Raindrop Optimization Algorithm to anticipate workload across successive prediction intervals. The proposed predictive model performs an iterative training on the historical data that further optimizes the workload traces for each prediction interval entailing an efficient balance between exploitation and exploration capabilities to attain quick convergence and enhanced prediction accuracy. The rigorous performance evaluation of proposed strategy has been done by utilizing the two benchmark data traces viz. Bitbrains and Google Cluster Dataset (GCD) over various prediction intervals. The results marked the substantial improvement in the predictive accuracy with iRD-NN outperforming other contemporary approaches, including, SVM, ARIMA, HHO and RNN-LSTM, attaining accuracy gains of 97.91%, 93.99%, 93.09% and 83.64%, respectively. Additionally, the proposed approach delivered a notable reduction in the error rate upto <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mn>1.6</mml:mn> <mml:mo>×</mml:mo> </mml:math> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:msup> <mml:mrow> <mml:mn>10</mml:mn> </mml:mrow> <mml:mrow> <mml:mo>−</mml:mo> <mml:mn>2</mml:mn> </mml:mrow> </mml:msup> </mml:math> accentuating its reliability and robustness. The results emphasized the model’s capability in augmenting resource allocation and operational competence in the cloud environment.