Abstract

The scheduling a set of periodic/sporadic real-time tasks on multiprocessor computing platform to achieve full system utilization with reduced overhead has been identified as a difficult problem and was well studied in the past. Proportional Fair (PFair) and Boundary Fair (BFair) are two optimal scheduling algorithms for discrete time multiprocessor systems. Here, PFair and its variants can handle both periodic and sporadic real-time tasks. Moreover, their performances have been recently evaluated in LITMUS. The original BFair scheduler cannot handle sporadic tasks as it relies on well-defined tasks' periods, which was recently extended for sporadic tasks (named as BF2). However, BFair scheduler has not been implemented and evaluated in real systems in the past. In this paper, we present an implementation of the extended BFair scheduler in Linux kernel 2.6.34 and discuss the lessons that we learned from the implementation. The implemented scheduler was evaluated on a ThinkStation with two Intel Xeon Quad-Core CPUs through extensive experiments. Our results show that, compared to the state-of-the-art optimal PFair scheduler (i.e., PD2, which is an early-release version of PFair) for sporadic tasks, BF2 has much less scheduling overhead with significantly reduced number of context switches and task migrations.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call