Abstract-Global real-time schedulers have earned the reputation of scaling poorly due to the high runtime overheads involved in global state management. In this paper, two mature implementations, one using fine-grained locking (SCHED DEADLINE) and one using coarse-grained locking (LITMUS RT 's G-EDF plugin), are evaluated and it is shown that, regardless of locking granularity, indeed neither scales well w.r.t. worst-case overheads due to excessive lock contention. To demonstrate that this is not an inherent limitation of global scheduling, the design of G-EDF-MP is presented, a global scheduler that uses message passing to avoid lock contention and cache-line sharing. It is shown to offer up to a 23-to 36-fold reduction in worst-case scheduling overhead on a 64-core platform, which translates into much improved schedulability (in some cases, more than 120 additional tasks can be supported).

This paper explores the problem of how to improve the average-case performance of real-time locking protocols, preferably without significantly deteriorating worst-case performance. Motivated by the futex implementation in Linux, where uncontended lock operations under the Priority Inheritance Protocol (PIP) do not incur mode-switching overheads, we extend this concept to more sophisticated protocols; namely the PCP, the MPCP and the FMLP + . We identify the challenges involved in implementing futexes for these protocols and present the design and evaluation of their implementations in LITMUS RT , a real-time extension of the Linux kernel. Our evaluation shows substantial improvements in the uncontended case (e.g., a futex implementation of the PCP lowers lock acquisition and release overheads by up to 75% and 92%, respectively), at the expense of some increases in worst-case overhead on par with Linux's existing futex implementation.

No abstract