Network Congestion Control via Route Cost Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network congestion control methods, such as DiffServ and RED, fail to effectively manage sudden increases in queue depth, leading to large-scale packet loss and failure to meet Quality of Service (QoS) parameters, especially in bursty and dynamic environments like Mobile Ad Hoc Networks (MANETs).
Innovation Solution
A method that determines the route cost metric for packets based on latency and discards packets if the metric exceeds the required QoS parameter, with a threshold set according to the normalized distribution of route cost metrics, allowing for biased packet discard during congestion, thereby reducing the number of packets not meeting QoS parameters and concentrating drops in less critical sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If DiffServ or RED congestion control methods are used, then network traffic can be managed with basic priority classes or random packet discarding, but large-scale packet loss occurs during sudden queue depth increases and QoS parameters are not met
Solution Approach 1:
The patent changes the parameter used for packet discarding from random selection (RED) or simple priority-based selection (DiffServ) to selection based on route cost metrics. Packets are discarded according to their route cost relative to a dynamically adjusted threshold, which changes based on queue depth and traffic conditions. This parameter change enables more intelligent packet selection that preserves QoS-critical traffic while discarding less important packets during congestion.
Solution Approach 2:
The patent performs preliminary assessment of packet route costs and QoS requirements before congestion fully develops. By calculating route cost metrics and comparing them against predicted thresholds based on queue depth trends, the system proactively discards packets unlikely to meet QoS requirements before they contribute to severe congestion. This preliminary action prevents the need for large-scale reactive packet discarding.
2Reliability
If packets are discarded randomly during congestion (RED policy), then queue overload is prevented, but packet loss is not selective and critical packets may be discarded while non-critical packets are retained
Solution Approach 1:
The patent applies local quality by treating different packets differently based on their individual route cost characteristics and QoS requirements. Instead of uniform random discarding (RED) or uniform priority treatment (DiffServ), each packet is evaluated against its specific route cost metric. Packets with higher route costs relative to the threshold are more likely to be discarded, while packets with lower route costs are preserved. This localized differentiation ensures critical packets are protected while non-critical packets are discarded during congestion.
3Reliability
If strict QoS thresholds are enforced for all packets, then packets meeting QoS requirements are preserved, but network utilization decreases and congestion control responsiveness is reduced
Solution Approach 1:
The patent implements dynamic threshold adjustment based on queue depth and traffic conditions. Rather than enforcing a fixed QoS threshold that would permanently limit network utilization, the discard threshold is dynamically adapted to current network state. When queue depth is low, the threshold allows more packets through to maximize utilization. When queue depth increases, the threshold adjusts to preserve QoS compliance. This dynamic behavior resolves the contradiction between strict QoS enforcement and network utilization.
Data Source
AI summary
A method and system of congestion control in a network are provided. A required quality of service (QoS) parameter, such as a maximum allowable latency, for a packet received at a queue in the network, and a route cost metric, such as accumulated and estimated latency, are determined, and the packet is either discarded if the route cost metric exceeds the required QoS parameter, or a discard bias value is set for the packet. Also, if the required QoS parameter exceeds the route cost metric, the method includes determining whether a congestion condition exists in the network, and if the congestion condition exists, biasing the packet for discard based on its latency if the route cost metric for the packet exceeds a threshold. The network may be an IP network, and a network such as a mobile ad hoc network (MANET).