Queueing Theory

Queueing theory studies systems where demand arrives over time and waits for service from limited capacity. Examples include packets waiting for a router, customers waiting for checkout, trucks waiting at a loading bay, parts waiting for a machine, support tickets waiting for technicians, or compute jobs waiting for processors.

The basic variables are arrival rate \lambda, service rate \mu, number of servers, queue discipline, buffer capacity, and variability of arrivals and service times. Utilization is often written:

where c is the number of parallel servers in a simple model. As utilization approaches one, waiting time can grow sharply even if average capacity appears sufficient.

Core relationships

Little’s Law is one of the most useful results:

where L is average number in the system and W is average time in the system for a stable process. It applies broadly and is useful for checking whether measured throughput, work-in-process, and lead time are consistent.

Queue models support staffing, buffer sizing, network capacity, service-level agreements, production-line design, hospital flow, call centers, maintenance crews, and logistics. When analytic assumptions are too simple, discrete-event simulation or Monte Carlo methods are used.

Common mistakes

A common mistake is designing for average demand only. Variability, burstiness, priority rules, setup time, downtime, blocking, and human behavior can dominate waiting time. Another is using utilization as the only performance metric while ignoring tail latency or service-level targets. A good review states arrival process, service-time distribution, queue discipline, capacity, downtime, abandonment or balking assumptions, and whether steady-state assumptions are valid.