Hierarchical Multiprocessor CPU Reservations for the Linux Kernel

This paper presents ongoing work in the development of a scheduling framework that will improve the ser-vice guarantees for soft real-time applications deployed on Linux. The scheduler has been designed around the current kernel infrastructure, trying to keep the changes minimal, and basing the scheduling policy on strong the-oretical results. The main goal is to achieve hierarchical distribution of the available computing power on multi-processor platforms, avoiding alterations to the existing user interfaces. The proposed framework exploits the hierarchical ar-rangement of tasks within groups and subgroups that is already possible within the Linux kernel. However, it adds the capability for each group to be assigned a pre-cise fraction of the computing power available on all the processors, using existing uniprocessor resource reser-vation techniques. Tasks are scheduled globally within each single group, and the partitions assigned to each group need not to be static, but can be dynamically bal-anced. Furthermore, the proposed mechanism can be used to support a variety of possible partitioning schemes using processor affinities. Also, the proposed implementation shows how to ad-dress the issue of predictable access to shared resources, reusing the priority inheritance infrastructure already present in the Linux kernel.