Deep learning-based multivariate resource utilization prediction for hotspots and coldspots mitigation in green cloud data centers

Abstract

Dynamic virtual machine (VM) consolidation is a constructive technique to enhance resource usage and is extensively employed to minimize data centers' energy consumption. However, in the current approaches, consolidation techniques are heavily relied on reducing the actively used physical servers (PMs) based on their current resource utilization without considering future resource demands. Also, many of the reported works for cloud workload prediction applied univariate time series-based forecasting models and neglected the dependency of other resource utilization metrics. Thus, resulting in inaccurate predictions, unnecessary migrations, high migration costs, and increased service level agreement violations (SLAVs) may nullify the consolidation benefits. To efficiently address this issue, we propose a multivariate resource usage prediction-based hotspots and coldspots mitigation approach that considers both the current and future usage of resources with O(sk) time complexity, where s and k denote the number of PMs and VMs, respectively. The proposed technique uses a clustering-based stacked bidirectional (Long Short-Term Memory) LSTM deep learning network to predict the future memory and CPU usage of PMs and VMs with high accuracy and O((Q(Q + W) * Θ) computational complexity, where Q, W, and Θ represent the number of hidden layer cells, outputs, and training epochs, respectively. Through extensive simulations based on Google's cluster workload traces, we demonstrate that our proposed method obtains substantial improvements in terms of prediction performance, energy-efficiency, actively used PMs, VM migrations, and SLA violations over the benchmark approaches.

MAG-D: A multivariate attention network based approach for cloud workload forecasting

“… RMSE CPU usage (In steps) Memory usage (In steps) Model All features 90 180 270 90 180 270 ARIMA [13] 0.00762 0.00773 0.00777 0.00535 0.00535 0.00536 Linear Regression [43] 0.00531 0.00535 0.00537 0.00424 0.00424 0.00425 GRU [43] 0.00539 0.00545 0.00547 0.00409 0.00411 0.00412 LSTM [38] 0.00533 0.00537 0.00539 0.00411 0.00413 0.00414 BiLSTM [20] 0.00539 0.00544 0.00547 0.00411 0.00413 0.00414 BiGRU [46] 0.00527 0.00534 0.00537 0.00390 0.00393 0.00395 CNN LSTM …”

“…This section entails discussing the performance of the various forecasting models at the CPU usage and Memory usage prediction tasks. In the current study, we have developed eight prediction models, namely ARIMA [13] , Linear Regression [43] , GRU [43] , LSTM [20] , [38] Bi-LSTM [43] , Bi-GRU [46] , CNN integrated LSTM (CNN LSTM) [40] and proposed Approach (MAG-D) for the target prediction tasks. Moreover, three variants of each model have been developed and trained covering different featural aspects of the input dataset, i.e.…”

“…Patel et al. proposed [43] , [44] multivariate forecasting-based hotspots and coldspots mitigation approach applying means clustering with stacked BiLSTM networks for cloud workload forecasting. To explore the applications of generative adversarial networks (GANs) for time-series forecasting, Yazdanian et al.…”

“…For estimating the overall computational complexity of proposed model, we adopted the fundamental model of neural network with cells of hidden layer, and number of inputs, and outputs respectively [43] , [46] . Now, the total parameters for a simple multi-layer perceptron (MLP) comprising a single hidden layer expressed as: .…”

A Robust Approach for Hotspots Prevention and Resolution in Cloud Services

“…They demonstrate that predictive-based scheduling increases utilization and prevents physical resource depletion. There are some other works, i.e., [10], [11], [12], [13], transforming the uncertainty of utilized capacity into a prediction problem via more complicated tools. However, they do not resolve the main challenge of the hotspot -the utilized capacity for VM changes frequently and irregularly over time -a prediction without a prescription cannot remove the hotspot permanently with minimal costs.…”