The pros & cons of queueing theory

Queuing Theory models queues. By modelling queues it aims to predict their behaviour and suggest strategies for mitigating the unpreparedness that causes them. Queues form when access to a resource is not granted with sufficient speed to satisfy all demand as it arises. No one likes queues, but they are better than the alternative -- denial of service.

Queues are the fairest way to deal with access. Alternative ordering strategies are heaps and stacks. In human situations, other alternatives are mobbing, or the “outcry” system. Neither of those options have the fairness of queuing. Queuing is a “FIFO” system. That means “first in, first out.” Stacks are “LIFO” -- “last in, first out.” Heaps are random. Queues for buffers in computing, they schedule work and they organise trains. They are at the heart of civilisation and create order. They are worth studying.

The study of queues and methods to deal with them aims at a perfect world where those who gain employment by studying clues get made redundant. Fortunately for queuing theorists, they will never be thrown on the dole queue, because there will always be queues. No business every plans their resources to meet peak demand. This strategy would leave large resources idle at non-peak periods and so would be a waste of an investment. Queues aim to even out peaks and troughs in demand by delaying services to requests arriving at peak times. The delay should be long enough to allow service to catch you with demand as demand declines. The equilibrium between resources and demand lies with resources being served at a constant rate, with demand managed to match that rate.

Queuing Theory is most concerned with the length of queues. In non-human situations, this means the facility for queues. For example, data packets arriving at a computer over a network may have to queue while the computer processes earlier arrivals. Queues in computing are not infinite; they have to be pre-set at a specific length. If the queue is not set long enough, packets that arrive when the queue is full will be rejected. On the other hand, sometimes it is better not to allow a queue to grow too long because wait times would be intolerable. At that point, the organisation should invest in improving processes to handle demand faster, which means buying new equipment.

Fine-tuning queues allows service providers to get away with providing too few resources. Sometimes, companies will hide behinds a supposed Queuing study to put off investing in increasing capacity and thereby they increase their profits. In human situations, Customer Services managers should not believe the results of a poorly conducted queuing study if it contradicts customer complaints of poor response times.