Abstract

Existing operating systems have remained mired in past paradigms, largely due to economic, as opposed to technical, reasons. Any of the assumptions upon which these systems were built have changed significantly over the past 20 years. The hardware on which operating systems run are dramatically different from their predecessors, due to the proliferation of multi-core CPUs, non-uniform memory architectures (NUMA), and hardware accelerators. Moreover, system deployments are markedly different (e.g., the increased adoption of containers for micro-services in shared-cloud environments and the proliferation of embedded systems through the adoption of IoT in the consumer device world as well as in mission critical systems such as transportation and the energy sector). These changes have far out stripped operating system evolution, which has been hampered by inertia around historical operating system design. As a consequence, existing operating systems suffer from several fundamental issues that lead to problems with performance, security, and software portability. Our work aims to address these issues in three ways: undertake a study of current operating system API usage by applications, build a green field OS (Zero) based on modern principles in a type safe language and transition components from the green field into existing operating systems as a way of supporting extensions in a principled way.