Skip to content
This is just a concept for a pressure based queue. Enqueueing becomes more difficult as the size of the queue increases.
Java Scala
Find file
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
src
.gitignore
README.md
build.sbt

README.md

A Concept for a Pressure-Based Queue

This is just a concept. I read a long while back about some guys that were using "pressure" to naturally throttle their systems, although I have no idea who did it... my memory has failed me. The idea is to make the act of queueing work take longer as the work backlog fills up.

Concurrency and asynchrony is awesome, except when the guys submitting the work can submit way faster than the guys doing the work can do it. For example, let's say your DB gets a huge amount of asynchronous requests and that makes your work slow down a bit. How do you slow down your clients?

The idea here is to take huge spikes in workload and smooth them out over time to allow for a more natural degradation of performance than brick walls of pain.

The source for the queue comes from Doug Lea's LinkedBoundedQueue and has been modified from there. This then gets used in a specific Akka dispatcher and we run an "eye test" to watch it throttle the input to the Actor.

Something went wrong with that request. Please try again.