Pedro Arantes
HomeBlogMe

At High Utilization Levels, Queues Grow Much Faster than They Can Shrink

At high utilization levels, losing capacity generates a queue much faster than regaining capacity can shrink.
#queueing-theory
Zettelkasten, January 6th, 2023.
The Network

Notes

At high utilization levels, losing capacity generates a queue much faster than regaining capacity can shrink. Remember that queues grow exponentially as you approach 100% of the queue capacity.

For example, suppose a single lane on a four-lane highway is blocked for 10 minutes in rush hour. In that case, it may create a 40 minutes traffic jam because losing 25 percent of capacity generates a queue much faster than regaining capacity can shrink it.

Questions:

  • What strategies could be applied to reduce queue growth at high utilization levels?
RecommendationsDo you want to see all posts instead?
Capacity Utilization Increases Queues Exponentially
Each time you halve the amount of excess capacity, you double the queue. You fold the queue as you move from 80 to 90 or 90 to 95 percent.
#queueing-theory
Quickly Apply Extra Resources to an Emerging Queue
To make supply-focused approaches effective, you must minimize the time between the onset of congestion and the application of added resources.
#high-state-queues
#queueing-theory
#supply-focused-approaches
#work-in-process
Control WIP by Shedding Requirements
When WIP becomes high, you can purge WIP by decreasing the batch size of your jobs. A way to achieve this is to shed the requirements of your jobs.
#cycle-time
#job-batching
#queueing-theory
#work-in-process
Block All Demand when WIP Reaches Its Upper Limit
You can block queues in two ways: 1. Eject the arriving job from the system. 2. Hold back the job in an upstream queue.
#product-development
#queueing-theory
#work-in-process
Cost of Delay
Cost of delay (COD) determines the financial impact on the company if it releases a product on the market earlier or later. When a product arrives on the market X months earlier, revenues also roll in X months earlier.
#cost-of-delay
#product-development
#queueing-theory