Network Congestion Detection via Jitter Spectrum Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing congestion control mechanisms in communications networks face challenges in accurately determining when congestion has ceased, leading to rate oscillations and inefficient link utilization, particularly affecting real-time media applications like voice and video.
Innovation Solution
A method in a receiving node that analyzes the frequency spectrum of time delays to differentiate between non-congestion and congestion-related frequency components, allowing for proactive rate adaptation by comparing reference and congestion jitter levels to determine the network's congestion state.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If packet loss is used as the main source for detecting congestion, then congestion control can be implemented, but real-time media applications like voice and video are affected because they are delay-sensitive and sensitive to packet loss
Solution Approach 1:
The patent introduces jitter (variation in packet arrival time) as an intermediary indicator to detect congestion. Instead of directly relying on packet loss, the system uses jitter measurements as a mediator that reflects network congestion conditions without directly impacting real-time media quality. This allows congestion detection while maintaining acceptable performance for delay-sensitive applications.
2Reliability
If TCP uses AIMD algorithm with decrease-to-half strategy at congestion, then congestion control is achieved, but the congestion window length is reduced significantly which may cause slow restoration after congestion
Solution Approach 1:
The patent implements dynamic rate adjustment based on real-time jitter measurements. Instead of using fixed AIMD parameters, the system dynamically modifies the congestion window adjustment strategy based on the current network conditions detected through jitter analysis. This allows for more flexible and adaptive rate control that can restore throughput faster while maintaining congestion control effectiveness.
3Ease of operation
If TCP-friendly Rate Control algorithms are used to smooth rate changes, then rate change intensity is reduced, but packet loss is emphasized as main congestion indicator which is not suitable for real-time media applications
Solution Approach 1:
The patent changes the fundamental parameter used for congestion detection from packet loss to jitter. This parameter change enables the system to maintain smooth rate adjustments while making congestion detection suitable for real-time media applications. The jitter-based approach provides continuous feedback that works well with rate smoothing algorithms without the drawbacks of packet loss-based methods.
4Reliability
If congestion is detected based on packet loss, then congestion control mechanisms can be triggered, but the application may have already stopped working when congestion is detected
Solution Approach 1:
The patent applies preliminary action by using jitter measurements that start changing before actual packet loss occurs. Since jitter increases as packets queue up in buffers before being dropped, the system can detect congestion conditions in advance and trigger rate control mechanisms proactively. This preliminary detection prevents the application from stopping work by enabling earlier intervention.
Data Source
AI summary
The invention relates to a method and arrangement for detecting congestion in a communications network (120). A receiving node (130) receives (710) data packets from a sending node (110) over a communications link (140). First and second parts (320, 330) of a frequency spectrum (310) of a time delay of the data packets are determined (720). The first part comprises non-congestion related low frequency components. The second part comprises congestion related and non-congestion related low frequency components when a load level of the communications network is above a threshold value (220) but substantially no congestion related low frequency components when the load level is below the threshold value. Then reference jitter level (yr) and congestion jitter level (yg) corresponding to the first and second parts respectively are estimated (740). Finally, congestion state information is determined (750) by comparing said congestion jitter level to said reference jitter level.


