News for hackers. Stuff that matters.
Re:Dice and Queues by jcalx on Wednesday April 09, 2025 @11:42PM (#43640412)
It's not explicitly mentioned in the article, but the arrival rate and queue sizes are related via Little's Law [0] — the average number of people in a system is equal to the arrival rate multiplied by the mean wait time, regardless of the arrival distribution, service distribution, or service order (e.g. FIFO, LIFO, random).

I was fortunate enough to take some great queueing theory classes in college, and fondly remember flailing about in Simio [1] trying to get a bank teller simulation to work. Really useful stuff to learn, although it did make me incredibly susceptible to Factorio and similar games.

[0] https://en.wikipedia.org/wiki/Little%27s_law

[1] https://www.youtube.com/watch?v=xCzAhKtucck

Re:Dice and Queues by dleeftink on Thursday April 10, 2025 @12:34AM (#43640637)
You may have seen it already, but the queueing saga by Defunctland remains fascinating [0].

[0]: https://youtu.be/9yjZpBq1XBE

Re:Dice and Queues by kqr on Thursday April 10, 2025 @12:40AM (#43640670)
Queueing theory is one of my favourite secret superweapons. You can use it to back-of-the-napkin solutions out of feasibility much quicker than it'd take to simulate.

I warmly recommend Harchol-Balter: https://www.amazon.com/Performance-Modeling-Design-Computer-... It is an eminintly practical book which is written in terms of computers, goes beyond the basic M/M/1 but without being overbearing.

(The other secret superpower is statistical process control.)

Re:Dice and Queues by code_biologist on Thursday April 10, 2025 @01:45AM (#43641021)
Agreed! The Principles of Product Development Flow by Donald G. Reinertsen[1] is one of my favorite product management books because it's very conceptual, but man the concepts hit home fast. This summary has a lot of great details [2] and the third chapter is entirely dedicated to the properties of queues.

After, Little's law, he introduces the First Queue Size Control Principle (Q13): Controlling queue size rather than capacity utilization offers a more effective way to manage cycle time and process efficiency. — No mister business stakeholder, adding items to the dev team's to-do list and making sure they are working 110% isn't going to get stuff delivered faster.

Diffusion Principle (Q15): Random processes can lead to queues spinning out of control; these high-queue states can last long and cause significant economic damage. — No mister business stakeholder, we can't fix this one little thing right now, it's bigger than you think and is going to make it harder to deliver the big important project.

[1] https://www.amazon.com/Principles-Product-Development-Flow-G...

[2] https://www.joecotellese.com/posts/principles-of-product-dev...

Re:Dice and Queues by JoeC3 on Saturday April 19, 2025 @02:41PM (#43738470)
It is a great PM book. Thanks for linking to my write up!
Re:Dice and Queues by vismit2000 on Friday April 11, 2025 @02:55AM (#43651186)
Little's law comes up in the beautiful book 'Programming Pearls' by Jon Bentley in 'Back of the Envelope' calculations:

- `Little’s Law`: The average number of objects in a queue is the product of the entry rate and the average holding time - Response-time formula for a multi-user system: - Assume n users of average think time z are connected to an arbitrary system with response time r. - Each user cycles between thinking and waiting-for-response, so the total number of jobs in the meta-system (consisting of users and the computer system) is fixed at n. - If you cut the path from the system's output to the users, you see a meta-system with average load n, average response time z + r, and throughput x (measured in jobs per time unit). - Little’s Law says n = x × (z + r), and solving for r gives r = n/x – z.