Building a Self-Healing Operating System (2007) [pdf] | Hacker News Reader

Building a Self-Healing Operating System (2007) [pdf]

choices.cs.illinois.edu

78 pointssteven7419 comments

The Moveable Feast Machine is a "robust-first" computing architecture designed to create vast distributed self-healing machines out of unreliable parts.

Here is a self healing city simulation in the Moveable Feast Machine. He causes a disaster in the middle of the city with the "nuke" tool, and it smoothly grows back and seamlessly repairs. He cuts holes in the wall around the city, and the grey goo molecular nanotechnology urban sprawl spreads out and takes over the entire world!

Robust-first Computing: Distributed City Generation: https://www.youtube.com/watch?v=XkSXERxucPc

Self-Healing Wire in the Movable Feast Machine: https://www.youtube.com/watch?v=EmPnjBhkbF0

A Robust Self-Healing Membrane in the Movable Feast Machine: https://www.youtube.com/watch?v=oq0uvF4mm7Y

Growth of self healing membranes in the Moveable Feast Machine: https://www.youtube.com/watch?v=L8zVYuV3AQw

Intercellular Transport in the Movable Feast Machine (Two cells are interconnected by self-healing wire): https://www.youtube.com/watch?v=6YucCpYCWpY

A NAND gate in the Movable Feast Machine (A NAND gate constructed of self-healing wire, using Message particles to represent quasi-electrons): https://www.youtube.com/watch?v=D5NVk12i3cM

The Movable Feast Machine is a robust, indefinitely scalable, computing architecture: https://movablefeastmachine.org/

I suppose we have seen advancements in this area in the past 10 years since this was published. Most applications and OS will at least attempt to recover all your work. Think a grey area still seems to be around significant upgrades to OSs when things can go seriously wrong. A recent upgrade of my Mac laptop collapsed spectacularly with much time spent on repairing the damage.

Good discussion on this topic https://www.compucom.com/blog/three-reasons-businesses-need-...

NixOS[0] makes all changes to your system atomically, and you can easily roll back to any previous state.

[0]: https://nixos.org/~eelco/pubs/phd-thesis.pdf

Right, applications have advanced...they are calling fsync (or similar expensive I/O operation to ensure consistent database), like, every time user clicks. These features are often poorly documented and understood, with huge overhead that IMO could be avoided if OS+apps were designed with consistency features in mind.

What would happen if the OS gradually forgot all changes that were made since factory settings, and all files that mattered were stored somewhere else? Would it only be annoying, or would it be comfortable to know that the system would gradually be heading toward the initial "clean" state? :D

I think you might be describing Windows 10. Resets default applications on each reboot.

I Googled a few more surveys for yall. Here they are:

A Survey on Self-Healing Systems: Approaches and Systems (2010)

http://www.infosys.tuwien.ac.at/Staff/sd/papers/Zeitschrift%...

Tools and Techniques for Designing and Evaluating Self-Healing Systems (slides)

http://www.cs.columbia.edu/~rg2023/pubs/Tools%20and%20Techni...

A Survey of Cloud Computing Fault Tolerance: Techniques and Implementation

http://tarjomefa.com/wp-content/uploads/2017/12/8351-English...

https://exif.tools/http://choices.cs.illinois.edu/selfhealin...

How about re-using / hacking intel's ME (Management Engine) for this type of fault tolerance?

Reply on news.ycombinator.com