Virtual Network Control via Demand Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual network embedding techniques face challenges in efficiently reconstructing virtual networks to meet changing demand, as current methods either overload the network with frequent node transfers or require lengthy computation times due to large-scale demand calculations.
Innovation Solution
A virtual network control apparatus with a physical network resource management agent and virtual network control agents that predict future demand and calculate resource allocation, minimizing node movement and computation time by sharing resource information across virtual networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If virtual network reconstruction is performed frequently to meet changing demand, then the virtual network can adapt to environmental changes, but the network load increases due to repeated transfer of image data
Solution Approach 1:
The patent applies preliminary action by predicting future demand before it actually occurs. The demand prediction unit forecasts future service demands, and the virtual network reconstruction is performed in advance based on these predictions, rather than reactively after demand changes are detected. This allows the system to prepare resource allocations proactively, reducing the frequency and intensity of actual reconstruction operations and thereby lowering network load from repeated image data transfers.
2Loss of energy
If intensive calculation of future demands for all virtual networks is performed to minimize node movement, then node transfer load is reduced, but computation time increases significantly
Solution Approach 1:
The patent applies segmentation by dividing the virtual network control into independent virtual network control agents, each responsible for a specific virtual network. Each agent independently performs demand prediction and resource allocation calculations for its assigned virtual network, rather than performing centralized calculations for all virtual networks. This segmentation reduces the computational burden on each agent, allowing frequent control operations without excessive computation time, while still achieving coordinated resource management across the entire system.
3Loss of time
If the control interval for virtual network reconstruction is lengthened to reduce computation time, then computation burden is reduced, but the system cannot follow environmental changes timely
Solution Approach 1:
The patent applies preliminary action by performing demand prediction and resource allocation calculations in advance, before environmental changes actually occur. The demand prediction unit forecasts future demand patterns, allowing the system to prepare reconstruction plans ahead of time. This proactive approach enables the system to maintain short control intervals and respond quickly to environmental changes, as the computational work has already been performed or is scheduled to be performed before the changes manifest.
4Ease of operation
If ad hoc reconstruction based on observed demand is performed, then the system can respond to current conditions, but virtual nodes transfer frequently causing large network load
Solution Approach 1:
The patent applies feedback by implementing a closed-loop control system where the demand prediction unit continuously monitors actual service demands, compares them with predicted demands, and uses this feedback information to refine future predictions. The resource allocation decisions are adjusted based on this feedback, allowing the system to respond accurately to current conditions while avoiding unnecessary reconstructions. This feedback mechanism reduces frequent virtual node transfers by making more accurate, informed decisions about when reconstruction is actually necessary.
Data Source
AI summary
A virtual network control apparatus configured to perform assignment of physical resources under management for a virtual network embedding request, including: a physical network resource management agent configured to manage use state of resources on a physical network at each future time; and a virtual network control agent, the virtual network control agent including: demand prediction means configured to perform prediction of future demand for a virtual network; and use resource determination means configured to calculate physical resources to be used at each future time by the virtual network based on a prediction result of the demand obtained by the demand prediction means and information of future use state of resources on the physical network obtained from the physical network resource management agent, and to notify the physical network resource management agent of information of physical resources of the calculation result.