Hacker News Reader

Comment by peterfirefly | Hacker News Reader

I expect one of the future frontiers in systems research (OS, compilers, etc) to be in this area: scaling things like file caches, prefetching, memory arenas, gc invocation (and gc types), thread pools, ...

They all need to be regulated, somehow, for best performance. Unfortunately, nobody really knows how to do that except in small isolated cases. What is even more unfortunate: very few of the people working in this area seem to know anything about control theory.

View on HN

i went down the rabbit hole on control theory when getting into highpower rocketry and electronics as a hobby. I suspect many of the concepts and algorithms have been independently discovered/implemented in high performance computing. It's all about trying to see through the noise and maintain a target in a dynamic environment. It maps pretty well to something like CPU utilization or IO performance or requests/second etc.

Do you have any good references to start learning about control theory? I'm after something for someone who already knows all the relevant mathematics but is unfamiliar with the applications.

There are be plenty of university level textbooks on control theory. For how to apply control theory to software problems there seems to be much less material though. Glyn Normington recommended the following book in another comment thread:

    Feedback Control for Computer Systems
    Philipp K. Janert
    330 pages
    O’Reilly (2013)
    ISBN: 978-1449361693

You'll find a little here and there within comp. sci. but the authors are usually not aware that they are doing control theory.

Typical things: scheduling algorithms for 1) processes, 2) disk reads/writes/seeks, TCP/IP stuff (slow start, throttling, window sizing).

For control theory viewed as part of engineering: "Control System Design" by Goodwin, Graebe, and Salgado.

There are many others but that was the one that worked best for me.

Start with this guy and then go as deep as you want with various textbooks.

https://youtube.com/@ControlLectures

At the end of the day, it's all an optimization problem. Even machine learning can be boiled down to being nonlinear optimization.

Reply on news.ycombinator.com