SQM-LRU: A Harmony Dual-Queue Management Algorithm to Control Non-Responsive LTF Flow and Achieve Service...

Area of Science:

• Computer Science
• Network Engineering
• Data Communications

Background:

• Increasing network application diversity necessitates advanced service differentiation.
• Non-responsive Long-Term Fast (LTF) flows are primary causes of congestion in Wide Area Networks (WANs).
• Existing queue management algorithms struggle to differentiate traffic effectively, hindering differentiated services.

Purpose of the Study:

• To propose SQM-LRU, an active queue management scheme for service differentiation in WANs.
• To detect and suppress non-responsive LTF flows to enable controllable delay and service differentiation.
• To satisfy delay requirements for responsive flows while maximizing throughput for non-responsive LTF flows.

Main Methods:

• Developed SQM-LRU utilizing Shadow Queue (SQ) and an improved Least-Recently-Used (LRU) strategy for flow detection.
• Implemented a dual-queue system with CoDel and RED for packet management based on flow detection.
• Classified flows into high or low priority sub-queues according to detection results.

Main Results:

• SQM-LRU demonstrated superior performance in flow detection compared to traditional methods.
• Achieved significant reductions in delay, jitter, and Flow Completion Time (FCT) for responsive flows.
• Reduced FCT by up to 50% and maintained 95% link utilization.

Conclusions:

• SQM-LRU effectively enables service differentiation by distinguishing responsive and non-responsive flows.
• The scheme is practical for real-world networks due to its low overhead and O(1) cost per packet.
• SQM-LRU offers a viable solution for managing network congestion and improving Quality of Service (QoS).