SDN Controller Network Statistics Estimation via Fluid Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network statistics estimation and prediction techniques face challenges in achieving real-time accuracy due to inherent propagation and processing delays associated with polling all routers, making it difficult to gather and correlate data in a timely manner for effective network management.
Innovation Solution
A method that uses a Software Defined Networking (SDN) controller to estimate and predict network statistics by polling only ingress border routers, utilizing network topology information and fluid modeling to calculate flow rates and queue dynamics for downstream devices without direct polling, enabling real-time network policy changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If polling data is collected from each router along the network path, then measurement precision of network statistics is improved, but loss of time increases due to propagation and processing delays
Solution Approach 1:
The patent extracts the polling function from all internal routers and concentrates it only at the border routers. The network controller polls only the border routers to obtain ingress and egress flow statistics, then uses fluid models to infer statistics for all internal routers along the path. This extraction eliminates the need to poll every router individually, thereby reducing time loss while maintaining measurement precision through mathematical modeling.
Solution Approach 2:
The patent introduces fluid models as an intermediary between the polled border router data and the unpolled internal router statistics. The fluid models act as a mediator that transforms the limited border router measurements into comprehensive network-wide statistics, allowing accurate estimation of internal router states without direct polling, thus resolving the time-precision contradiction.
2Productivity
If aggressive polling of all routers is performed to achieve real-time statistics, then productivity of network monitoring is improved, but device complexity increases
Solution Approach 1:
The patent extracts the polling burden from the entire router population and concentrates it on border routers only. This reduces the complexity of the polling system by eliminating the need to manage polling relationships with every internal router, while maintaining high productivity through efficient use of border router data and fluid model calculations.
Solution Approach 2:
The border routers serve multiple functions: they act as both the polling targets for the network controller and as data sources for inferring the states of all internal routers. This multi-functionality reduces overall system complexity by having fewer devices perform multiple roles, eliminating the need for a complex distributed polling architecture.
3Loss of time
If polling frequency is increased to reduce time delay, then loss of time is reduced, but use of energy increases due to aggressive polling
Solution Approach 1:
The patent extracts the polling activity from all routers and concentrates it on border routers only. This dramatically reduces the total energy consumption of the polling system while maintaining real-time monitoring capability, because the fluid models can infer internal router states continuously from the border router data without requiring additional polling energy.
Solution Approach 2:
The fluid models enable the system to self-generate comprehensive network statistics from the limited border router polling data. Instead of requiring energy-intensive polling of all routers, the models automatically compute internal router states from the border router measurements, making the system energy-efficient while maintaining real-time accuracy.
Data Source
AI summary
A network computing device determines a network topology for at least one network flow path between at least one ingress network border device and at least one egress network border device. The network computing device receives a message containing data indicating flow statistics for the at least one ingress network border device. The network computing device generates flow statistics for at least one network device along the at least one network flow path from the network topology and the flow statistics for the at least one ingress network border device. The network computing device generates the flow statistics for at least one network device along the at least one network flow path without receiving flow statistics from the at least one network device along the at least one network flow path.