Per-Flow Call Admission Control via Predictive Tunnel QoS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current call admission control (CAC) approaches in SD-WANs are limited as they make all-or-nothing eligibility decisions for tunnels based on static performance characteristics, failing to account for existing traffic flows and resource consumption, leading to inefficient routing of traffic.
Innovation Solution
Implementing a predictive model using machine learning to dynamically assess whether a tunnel can satisfy the service level agreement (SLA) of a new traffic flow, allowing for per-flow call admission control and routing decisions based on real-time performance predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If static, preconfigured tunnel capabilities are used for CAC decisions, then the CAC process is simple and fast, but the routing decisions are inaccurate and do not reflect real-time tunnel performance
Solution Approach 1:
The patent transforms the static CAC approach into a dynamic one by continuously monitoring tunnel performance metrics (latency, jitter, packet loss, bandwidth utilization) and using these real-time measurements to assess tunnel capability. This allows the system to adapt to changing network conditions while maintaining accurate routing decisions.
Solution Approach 2:
The patent implements a feedback mechanism where tunnel performance is continuously measured and fed back to the CAC system. This feedback loop enables the system to update its understanding of tunnel capabilities dynamically, improving measurement precision without requiring complete system redesign.
2Productivity
If all-or-nothing tunnel eligibility decisions are made, then the CAC decision process is simple, but traffic routing efficiency deteriorates as tunnels are either fully eligible or completely ineligible
Solution Approach 1:
The patent applies local quality by making CAC decisions specific to each traffic flow rather than applying a uniform eligibility decision to all traffic. Each flow is evaluated individually against tunnel capabilities, allowing fine-grained optimization where different flows can use different tunnels based on their specific requirements and current tunnel conditions.
Solution Approach 2:
The patent segments the CAC decision process into per-flow evaluations rather than treating all traffic as a single unit. This segmentation allows the system to optimize routing for each flow independently, improving overall traffic routing efficiency by matching specific flows to most suitable tunnels.
3Reliability
If existing traffic flows and resource consumption are not considered, then the CAC decision process is simple, but tunnel overload or underutilization occurs leading to poor QoS
Solution Approach 1:
The patent uses feedback from continuous monitoring of existing traffic flows and resource consumption metrics. This feedback enables the system to assess current tunnel utilization and adjust CAC decisions accordingly, preventing both overload and underutilization while maintaining reliable QoS guarantees.
Solution Approach 2:
The patent performs preliminary assessment of tunnel capacity by considering existing flows and resource consumption before making new flow admission decisions. This preliminary action ensures that QoS requirements can be met by evaluating current resource availability upfront, rather than reacting to congestion after it occurs.
Data Source
AI summary
In one embodiment, a device identifies a new traffic flow in a network. The device determines a service level agreement (SLA) associated with the new traffic flow. The device uses a machine learning model to predict whether a particular tunnel in the network can satisfy the determined SLA of the traffic were the traffic flow routed onto the tunnel. The device performs call admission control to route the new traffic flow onto the particular tunnel, based on a prediction that the tunnel can satisfy the determined SLA of the traffic.