Home
mOS for HPC is an operating systems research project at Intel Corp., targeting extreme scale HPC systems. It aims to deliver a high performance computing environment with the scalability, low noise, and repeatability expected of lightweight kernels (LWK), while maintaining overall Linux compatibility that HPC applications need.
mOS for HPC remains under development at this time. These materials are being made available to interested parties to explore, to test drive, and to provide feedback through the mailing list. Consider the quality level to be pre-alpha and as-is. It is not intended to be used in production or for business critical uses. Installing mOS requires some systems software knowledge; namely the ability to build and install a Linux kernel. (The mOS LWK is embedded in a Linux kernel.) Support is limited by the development team's ability to respond through the mailing list.
We are always interested in hearing from potential users and collaborators. Please contact us at: [email protected]
Please see the appropriate version of readme, administrator's guide, and user's guide in the document list at the right. These pages contain information about installing and using mOS.
The book Operating Systems for Supercomputers and High Performance Computing contains a chapter (18) about the design of mOS. Other chapters explain how lightweight kernels came to be and the evolution, via Linux, toward multi-kernels like mOS. It is available from Springer.
Together with our McKernel collaborators at RIKEN, we have written and published several peer-reviewed papers. Performance and Scalability of Lightweight Multi-kernel Based Operating Systems, 2018 IEEE International Parallel and Distributed Processing Symposium (IPDPS) Publication link. This paper shows some of the first scaling results from McKernel and mOS. Both kernels have evolved considerably since these measurements were made, but the graphs and explanations in this paper show where we are headed.
The original idea for mOS was that it would support large-scale, traditional MPI+x, BSP style applications. Most typically simulations of physical and chemical systems. As time went on, it became clear that mOS needs to support a broader set of applications and modern technologies. This paper looks at one such aspect: How multi-kernels can support containers. Toward Full Specialization of the {HPC} Software Stack: Reconciling Application Containers and Lightweight Multi-kernels, Proceedings of the 7th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS '17) Publication link.
Several teams around the world are working on multi-kernels for large-scale HPC systems. This paper provides a list and comparison of what our collaborators and colleagues are doing. A Multi-Kernel Survey for High-Performance Computing, Proceedings of the 6th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS '16) Publication link.
In 2015 we started to write code. Most of it is gone now, but we started to have a clearer picture of what mOS should be and how it could be implemented. This paper explains our thinking back then. Exploring the Design Space of Combining Linux with Lightweight Kernels for Extreme Scale Computing, Proceedings of the 5th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS '15) Publication link.
The first mOS paper. Here we explain some of the early concepts. Not all the ideas in this paper made it into the implementation and some concept that survived have transformed through the design and implementation stages of this project. mOS: An Architecture for Extreme-scale Operating Systems, Proceedings of the 4th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS '14) Publication link.
