QoS Priority Queuing
Managing Congestion in Converged Networks
The Necessity of Quality of Service
In a best-effort network, all traffic is treated equally. However, real-time applications like VoIP and Video conferencing are sensitive to Jitter and Latency, while file transfers are only sensitive to throughput. QoS allows us to classify and prioritize these streams.
Fundamental Queuing Algorithms
1. FIFO (First-In, First-Out)
The default. Packets departure in the order they arrived. No prioritization.
2. Priority Queuing (PQ)
Packets are placed into strictly prioritized queues (High, Medium, Normal, Low). The High queue must be completely empty before any packet in the Medium queue is sent.
3. Weighted Fair Queuing (WFQ)
Automatically divides bandwidth among active flows based on weight. Lower-volume flows (like ICMP or VoIP control) get better prioritization than high-volume TCP streams.
4. Low Latency Queuing (LLQ)
A hybrid of PQ and CBWFQ. It adds a "Priority" class to a weighted queuing system, ensuring that real-time voice packets are always serviced first without starving other classes entirely.
Mathematics of Delay
Serialization delay is the time required to place bits onto the physical medium.
On slow links, a large 1500B data packet can block a 64B VoIP packet for several milliseconds, causing jitter. QoS mitigates this by allowing the scheduler to jump the queue.
Conclusion
Effective QoS design requires identifying 'Mission Critical' traffic and ensuring it has a reserved 'slice' of the bandwidth during periods of heavy congestion.