K42: Lessons for the OS community

Abstract

We started the K42 project more than ten years ago with the ambitious goal of developing an operating system for next-generation hardware that would be widely valued and thus widely used. Based on the premise that current operating systems were not designed to be scalable, customizable, or maintainable, we set forth to rectify that by applying proven techniques from other disciplines to operating systems and by developing additional innovative mechanisms. Now, ten year later, K42 is used by ten or so universities and national labs for research purposes, not ten million information technology departments desiring better everyday computing platforms. As a presentation to the primary operating systems community we provide an examination from two different perspectives as to what went right and what went wrong. First, we concentrate on what technology worked well and why, and what technology failed or caused undue difficulties, and why. Second, based on that experience, we provide our thoughts on the state and direction of the OS community at large. To be clear, this paper is neither a results paper nor an overview paper; we refer to other papers for background material. Rather, it is an exploration by researchers with experience with at least six different previous operating systems of the merit of technologies investigated in K42, and an extrapolation of the implications of that experience to the wider operating system community.