Cloud Provisioning Accelerator Using Ensemble Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud computing technologies face challenges in providing instant virtual machine (VM) provisioning due to latency issues, with state-of-art VM provisioning taking tens of minutes, which is unacceptable for urgent computing needs, and existing prediction techniques fail to accurately forecast dynamic VM demands.
Innovation Solution
The solution involves extracting high-level characteristics from VM provisioning request streams, generating prediction models for different VM types, and applying a regression-correlation ensemble to predict future VM demands, utilizing a novel measurement called Cloud Prediction Cost to guide the prediction process and reduce provisioning time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional VM provisioning technology is used, then VMs can be provisioned with standard latency, but provisioning time takes tens of minutes which is unacceptable for urgent computing needs
Solution Approach 1:
The system performs preliminary actions by predicting future VM demands using machine learning models and pre-provisioning VMs before they are actually needed. This allows the system to have VMs ready in advance for urgent computing needs, reducing provisioning time from tens of minutes to seconds when demands are predicted accurately.
Solution Approach 2:
The system dynamically adjusts provisioning strategies based on predicted demand patterns. It uses regression-correlation ensemble models that adapt to changing workload characteristics, allowing the provisioning system to optimize between pre-provisioning and on-demand provisioning based on real-time and historical data patterns.
2Measurement precision
If existing prediction techniques are used, then some forecasting capability is provided, but they fail to accurately forecast dynamic VM demands
Solution Approach 1:
The system merges multiple prediction models into a regression-correlation ensemble that combines different forecasting approaches. This ensemble methodology integrates various machine learning models to predict VM demands, improving overall prediction accuracy by leveraging the strengths of different models and handling diverse demand patterns more effectively.
Solution Approach 2:
The system implements feedback mechanisms where prediction accuracy is continuously evaluated and models are retrained using actual demand data. This feedback loop allows the prediction system to learn from past performance and improve its forecasting capability over time, adapting to changing workload patterns and seasonal variations in VM demands.
Data Source
AI summary
High level characteristics are extracted from a virtual machine provisioning request stream for a cloud computing environment. A plurality of prediction models are generated for a plurality of different types of virtual machines. A regression-correlation ensemble is applied to appropriate ones of the prediction models to predict future virtual machine demand for the cloud computing environment.