Cloud Infrastructure Estimation and Auto-Scaling Using Recurrent Cartesian Genetic Programming-Based ANN

Abstract

Use of cloud resources has increased with the increasing trend of organizations and governments towards cloud adaptation. This increase in cloud resource usage, leads to enormous amounts of energy consumption by cloud data center servers. Energy can be conserved in a cloud server by demand-based scaling of resources. But reactive scaling may lead to excessive scaling. That, in turn, results in enormous energy consumption by useless scale up and scale down. The scaling granularity can also result in excessive scaling of the resource. Without a proper mechanism for estimating cloud resource usage may lead to significant scaling overheads. To overcome, such inefficiencies, we present Cartesian genetic programming based neural network for resource estimation and a rule-based scaling system for IaaS cloud server. Our system consists of a resource monitor, a resource estimator and a scaling mechanism. The resource monitor takes resource utilizations and feeds to the estimator for efficient estimation of resources. The scaling system uses the resource estimator's output for scaling the resource with the granularity of a CPU core. The proposed method has been trained and tested with real traces of Bitbrains data center, producing promising results in real-time. It has shown better prediction accuracy and energy efficiency than predictive scaling systems from literature. INDEX TERMS Artificial neural networks, auto-scaling, cartesian genetic programming, energy efficiency, evolutionary computation, green computing, infrastructure as service, workload prediction, cloud server.

Scaling UPF Instances in 5G/6G Core With Deep Reinforcement Learning

“…Figure 8 shows us a comparison between the evaluation of a trained DQN and a trained PPO agent. We can see that while the PPO could adapt well to the varying arrival rate in (16) and found a balance between the Pod count and the blocking rate, the DQN could not keep the Pod count as low and overprovisioned the Pods. We also have to note that the DQN seemed to be much more unstable as in many cases it failed to even learn a policy that would adapt to traffic changes, whereas the PPO algorithm could find a good policy throughout every run.…”

“…In a hybrid approach, Gervásio et al [15] combined an ensemble of prediction models with a dynamic threshold algorithm to scale virtual machines in an AWS cloud. Ullah et al [16] used genetic algorithm and artificial neural networks to predict CPU usage in a cloud and then used a thresholdbased rule.…”

“…The measured arrival rate λ during the evaluation phase when traffic was generated according to λ eval in(16).…”

Hybrid Auto-Scaled Service-Cloud-Based Predictive Workload Modeling and Analysis for Smart Campus System

“…In the parametric frameworks, it is linear in the shape of ARMA, Auto Regressive Moving Average. Without an appropriate procedure to predict, the use of cloud resources may face a scalability problem [28]. To handle the scaling situation, Neural Network's approach based on the programming of cartesian genetic to estimate resources scalability based on the rule for a cloud server.…”

A Queueing Model for Threshold-Based Scaling of UPF Instances in 5G Core

“…Gervásio et al [20] in their hybrid autoscaling proposal, combined a self-adaptive prediction and reactive approach for optimal configuration of the threshold values for scaling operations. Ullah et al [21] presented the combination of predictive and reactive approach where Cartesian genetic programming based neural network is used for resource estimation and a rule-based scaling.…”