NFV Resource Forecasting for Physical-to-VM Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing NFV infrastructure management systems face issues of resource underutilization and misalignment with actual load conditions, leading to operational burdens and performance disruptions due to frequent reallocations and overprovisioning, as they rely on static information and short-term forecasts without considering physical resource availability.
Innovation Solution
A method using an automatic learning mechanism, such as advanced statistical or machine learning techniques, to forecast resource requirements over a medium/long-term time frame, enabling optimal allocation of physical resources to virtual machines based on historical data and business forecasts, minimizing maintenance operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If static information and heuristic allocation algorithms are used to allocate virtual machines, then the allocation process is simple and fast, but resource utilization is poor and misalignment with actual load conditions occurs
Solution Approach 1:
The system performs preliminary actions by using automatic learning mechanisms to forecast future resource requirements before actual load conditions occur. Historical data is analyzed in advance to predict future needs, allowing the system to proactively allocate resources rather than reacting to static specifications or real-time demands alone.
Solution Approach 2:
The system implements feedback loops where actual resource consumption data is continuously collected and fed back into the automatic learning mechanism. This feedback enables the model to refine its predictions and improve allocation accuracy over time, adapting to changing load patterns and optimizing resource utilization dynamically.
2Reliability
If overprovisioning policies are applied to compensate for resource underutilization, then service reliability is improved, but infrastructure capacity is wasted and operational burden increases
Solution Approach 1:
The system transitions from static overprovisioning policies to dynamic resource allocation based on real-time and forecasted load conditions. The automatic learning mechanism continuously adapts allocation decisions to actual usage patterns, allowing the infrastructure to expand or contract capacity as needed rather than maintaining fixed overprovisioned levels.
Solution Approach 2:
The system changes the parameter of allocation timing from immediate or static to forecast-based. By predicting future resource requirements, the system can optimize the timing and amount of resource allocation, reducing unnecessary infrastructure capacity while ensuring reliability through proactive resource preparation.
3Measurement precision
If frequent reallocations and infrastructure resizing operations are performed, then resource allocation accuracy is improved, but operational burden increases and performance disruption risk increases
Solution Approach 1:
The automatic learning mechanism performs preliminary analysis of historical data to forecast future resource requirements, enabling accurate allocation decisions to be made in advance. This reduces the frequency of reactive reallocations and resizing operations, as the system can proactively adjust resources before load conditions change.
Solution Approach 2:
The system implements dynamic allocation strategies that adapt to changing conditions without requiring frequent manual interventions. The automatic learning mechanism continuously optimizes allocation decisions based on actual usage patterns, reducing operational burden while maintaining high allocation accuracy through automated rather than manual processes.
Data Source
Figure 1
AI summary
A method of managing resources of an infrastructure for Network Function Virtualization comprising the steps of: providing (1) a first database containing historical telephony infrastructure management data; providing (2) a control unit in signal communication with the database; providing (3) an automatic learning mechanism; extrapolating (4), by the control unit, at least one first subset of historical data related to a predetermined first time window from the historical telephony infrastructure management data; extrapolating (5), by the control unit, at least one second subset of historical data related to a predetermined second time window from the historical telephony infrastructure management data; defining (6), by the control unit, a historical input vector comprising the historical data of the first subset of historical data; defining (6) by means of the control unit a historical input vector comprising the historical data of the first subset of historical data; defining (7), by the control unit, a historical output vector comprising the historical data of the second subset of historical data; training (8), by the control unit, the automatic learning mechanism to obtain, at its output, the historical output vector by providing the historical input vector at the input of the automatic learning mechanism; acquiring (9), by the control unit, at least one subset of current telephony infrastructure management data related to a current time window; defining (10) a historical input vector comprising the subset of current data; defining (10) a current input vector comprising the subset of current data; providing (11) the current input vector at the input of the automatic learning mechanism, which is trained to obtain a prediction output vector; analyzing (12) the prediction output vector for defining the telephony infrastructure resource consumption; providing (13) an optimal resource allocation algorithm; executing (14) the optimal resource allocation algorithm by providing the telephony infrastructure resource consumption at its input to obtain an optimization vector at the output, representative of the number of physical resources to be allocated to the virtual machines and the resources committed by each virtual machine for optimized management of telephony infrastructure resources