Yes, there have been functional demonstrations of an Exokernel, though widespread adoption remains limited.
Yes, there have been functional demonstrations of an Exokernel, though widespread adoption remains limited.
I'm curious about whether this kind of kernel has already been tested outside academic settings. From what I understand, they tend to be designed with minimal layers of abstraction, making them highly efficient. If true, how would a basic program interact with hardware? Would it need to communicate directly with drivers for each operation? For multithreading, would POSIX threads provide a similar alternative? Theoretical performance gains are interesting—an article from Cornell mentioned a research kernel achieving faster IPC than L4, around 30 vs 50 instructions per cycle. That’s quite remarkable given L4 was originally written in assembly for maximum speed. It still fascinates me, though modern security concerns might limit practical use of such an OS.
Zircon offers unique features but functions as a microkernel emphasizing isolation and least privilege, though it relies heavily on system calls for practicality. For future readers, the only viable exokernel I discovered is a branch from KolibriOS—an extremely lightweight OS with a 100kb kernel written in pure assembly. It hasn't seen updates in over a year and was originally designed for ultra-embedded systems.