The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Workshop:11th International Workshop on Runtime and Operating Systems for Supercomputers ROSS
Authors: Kyle Hale (Illinois Institute of Technology) and Simone Campanoni, Nikos Hardavellas, and Peter Dinda (Northwestern University)
Abstract: The layered structure of the system software stacks we use today allows for separation of concerns and increases portability. However, the confluence of widely available virtualization and hardware partitioning technology, new OS techniques, rapidly changing hardware, and significant advances in compiler technology together present a ripe opportunity for restructuring the stack, particularly to support effective parallel execution. We argue that there are cases where layers, particularly the compiler, run-time, kernel, and hardware, should be interwoven, enabling new optimizations and abstractions. We present four examples where we have successfully applied this interweaving model of system design, and we outline several lines of promising ongoing work.