Turning a Highly Reliable, Self-Repairing μ-Kernel OS into a Multicore-Based OS

Nowadays multicore processors have gradually spread. Accordingly, operating systems (OSs) have to follow up. Since multicore failures affect software systems, most schedulers distribute tasks over several dies to reduce recovery time in case of single die failure, instead of exploiting high-speed links between cores on one die. To fully exploit multicore features within a die, a highly reliable, fault-tolerant OS has to be developed. This maximizes the performance on a single die and on multiple dies. Microkernel-based OSs are suitable to implement the necessary extensions to support multicore architectures. MINIX3 is a highly reliable, self-repairing multiserver microkernel OS, but it doesn't support multicore architectures. This paper presents a proposed design and implementation of modifications to MINIX3 to support multicore architectures. The modified MINIX3 has been evaluated considering three issues: reliability, multicore port performance, and contention on the locking scheme.