SDN Traffic Congestion Control via Stability Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network management systems struggle to effectively control congestion in software-defined networks (SDNs) that handle both light tail and heavy tail traffic, leading to delays and inefficiencies due to their reliance on heuristic or measurement-based designs that do not account for hybrid traffic patterns.
Innovation Solution
A fast-adaptive congestion-free traffic engineering mechanism that calculates end-to-end delays and determines network stability thresholds to adjust traffic rates and routing policies, using techniques like last in first out (LIFO) delay maximum weight scheduling and alternating direction method of multipliers (ADMM) to optimize traffic flow and prevent congestion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If light tail traffic is controlled by adjusting delay based on queue length in switches or by packet delays at the head of the line, then loss-free delivery is achieved for purely LT traffic, but E2E delay increases for hybrid traffic that includes both LT and HT traffic
Solution Approach 1:
The patent changes the control parameter from simple queue length-based delay adjustment to a more sophisticated admission control mechanism that uses stability regions. The controller determines whether to accept offered traffic loads by comparing them against dynamically calculated stability thresholds, which are functions of current network state and traffic characteristics. This parameter change allows the system to maintain loss-free delivery for LT traffic while avoiding excessive E2E delay for hybrid traffic by making informed admission decisions before congestion occurs.
Solution Approach 2:
The patent implements preliminary action through proactive admission control at the controller. Instead of reacting to congestion after it occurs, the system calculates stability regions and thresholds in advance and uses them to make admission decisions before traffic enters the network. This preliminary assessment allows the system to prevent congestion buildup that would cause E2E delay, while still accepting LT traffic that can be delivered loss-free.
2Device complexity
If heuristic or measurement-based designs are used for congestion control, then implementation is simpler, but network stability and performance deteriorate due to lack of account for hybrid traffic patterns
Solution Approach 1:
The patent implements feedback through a closed-loop control mechanism where the controller continuously monitors accepted traffic statistics and uses this information to recalculate stability regions and thresholds. This feedback loop allows the system to adapt to changing network conditions and traffic patterns in real-time, maintaining network stability even as traffic composition changes. The controller adjusts admission decisions based on actual network state, creating a self-regulating system that handles hybrid traffic patterns effectively.
Solution Approach 2:
The patent introduces dynamics by making the admission control thresholds and stability regions time-varying rather than static. The system dynamically calculates these parameters based on current traffic statistics, network load, and traffic type compositions. This dynamic adaptation allows the network to respond to changing conditions without requiring complex predetermined rules, achieving both simplicity in implementation and reliability in performance through automated adaptive behavior.
3Device complexity
If rate or window size assignments are used for simple admission control, then control mechanism is simpler, but signaling overhead increases due to exchange of queue length information between switches
Solution Approach 1:
The patent extracts the admission control function from the individual switches and consolidates it at the centralized controller. Instead of each switch independently managing its own queue length information and exchanging data with neighbors, the controller receives traffic statistics, calculates stability regions, and makes unified admission decisions. This extraction of the control function to the controller level reduces signaling overhead between switches while maintaining simple implementation at the edge devices.
Solution Approach 2:
The patent applies universality by having the controller perform multiple functions: collecting traffic statistics, calculating stability regions, determining admission decisions, and providing rate assignments. This multi-functional approach consolidates what would otherwise require multiple separate control mechanisms and communication protocols into a single unified system, reducing overall signaling overhead while maintaining effective congestion control.
Data Source
AI summary
A method includes obtaining traffic statistics of accepted hybrid traffic at a controller of a software defined network that includes multiple local switches coupled by links, calculating an end to end delay associated with the accepted hybrid traffic, determining network stability thresholds from each local switch as a function of network congestion, determining an adjusted rate decision policy as a function of the end to end delay and the network stability thresholds, the adjusted rate decision policy for use by an edge switch to determine whether or not to accept offered traffic loads.