Operating Systems Research

Operating systems research, like in most things, falls into two categories. Finding new techniques to accomplish known worthy goals, and finding new worthy goals to accomplish. Respectively, practical research and basic research.

Research is distinguished from OperatingSystemsDesign, which is conceiving of the concepts which reconcile the chosen goals within the limits of the technology. Which is in turn distinguished from OperatingSystemsImplementation.

OS Research is almost completely dead. And what precious little research exists is almost all practical. Even research into very basic questions tends to be completely perverted by practical considerations. For instance, Rosenblum's basic research into log-structured filesystems was focused on their read-write performance characteristics instead of their novel properties.

• Which novel properties, you mean the database-like ACID guarantees?

• If properly implemented, an LFS is more portable and dynamic than a traditional filesystem. You would be able to transport the log to another medium, or change its size within that medium, without any difficulty. And that's nothing compared with the ability to maintain an ordered time sequence of globally consistent state. This would let you revert the system backwards to any point in its past, without any prior arrangement.

Only research into HumanComputerInteraction seems to be somewhat immune to this corruption effect.

As far as basic research into operating systems, something that might hope to uncover new OperatingSystemsDesignPrinciples, the field is dead. Not even zombie undead but dead dead.

• See RobPike's "Systems Software Research is Irrelevant" (feb 2000) http://www.eng.uwaterloo.ca/~ejones/writing/systemsresearch.html

Part of the reason why basic OS research is so dead is the fact that OperatingSystemDesigners are about 30 years behind on the research.

By that I assume you mean there's still 30 years worth of practical research to incorporate into working systems before we do more basic research?

There's still 30 years of basic research to incorporate into working systems, nevermind the practical research.

---

CategoryOperatingSystem CategoryResearch