Queueing Theory and Congestion Fundamentals
Queues as the Core Model of Networking
Why queueing dominates performance
Packets contend for finite transmission resources at routers, switches, wireless schedulers, and end hosts. When arrival rate exceeds service rate, queue length grows, delay increases, and loss may occur if buffers overflow.
Queueing vocabulary
Stability and delay
For many simple queues, stability requires . As utilization approaches , average delay rises sharply. This is one reason highly utilized networks often exhibit bursty latency even when average throughput looks healthy.
Phenomena and causes
| Phenomenon | Typical cause |
|---|---|
| Bufferbloat | Excessively large queues causing high latency |
| Tail drop | Dropping packets only when a buffer fills |
| Head-of-line blocking | A blocked packet delaying others behind it |
| Queueing collapse | Unstable feedback between congestion and retransmission |
What condition is typically required for a simple queue to remain stable?
In basic queueing models, stability generally requires arrival rate below service rate, i.e. utilization below 1.
Correct answer:
State Little's law in words.
Little's law is widely applicable across queueing systems and networking models.
Correct answer: The average number of items in a stable system equals the arrival rate multiplied by the average time an item spends in the system.
Key insight
High throughput does not imply low delay. In fact, latency can rise dramatically near saturation even when packet loss is still modest.
Queueing analysis workflow
-
1
Identify the service discipline and arrival process.
-
2
Estimate utilization and identify bottlenecks.
-
3
Check whether the system is stable.
-
4
Predict delay and loss trends under load.
-
5
Validate model predictions against measurements or simulations.
What is a classic symptom of bufferbloat?
Bufferbloat refers to inflated latency caused by oversized buffers and persistent queues.
Correct answer: Excessive latency due to large queues