Adaptive Resource Utilization Prediction System for Infrastructure as a Service Cloud.

Area of Science:

• Cloud Computing
• Resource Management
• Predictive Analytics

Background:

• Infrastructure as a Service (IaaS) provides scalable compute, network, and storage resources.
• Effective resource usage prediction is crucial for dynamic scaling, cost efficiency, and energy conservation in cloud environments.
• Accurate demand forecasting ensures Quality of Service (QoS) maintenance.

Purpose of the Study:

• To present a novel real-time resource usage prediction system for IaaS clouds.
• To enable efficient dynamic scaling and resource management through accurate demand forecasting.
• To optimize cloud operations for cost and energy efficiency while preserving QoS.

Main Methods:

• Real-time resource utilization data is collected and buffered based on resource type and time span.
• Statistical analysis determines if buffered data follows a Gaussian distribution.
• Autoregressive Integrated Moving Average (ARIMA) or Autoregressive Neural Network (AR-NN) models are applied based on data distribution, with model selection via AIC or NIC criteria.

Main Results:

• The developed system was evaluated using real CPU utilization traces from an IaaS cloud environment.
• The system demonstrated the capability to predict resource usage patterns effectively.
• The adaptive model selection (ARIMA/AR-NN) based on data distribution proved effective.

Conclusions:

• The proposed real-time prediction system enhances IaaS cloud resource management.
• Accurate resource usage prediction facilitates optimized dynamic scaling, leading to cost and energy savings.
• The system provides a robust solution for maintaining QoS in fluctuating cloud environments.