Improving Reliability of Multi-/Many-Core Processors by Using NMR-MPar Approach

2019 
The new trend in computing systems is providing solutions by using multicore and many-core processors. COTS processors are preferred because they offer a high performance with low-power consumption within an affordable price. Lately these devices have been used in High Performance Computing systems due to their massive parallelism and low-power budget. For the last decade, industrial and academic partners have worked together to overcome with dependability issues to extend their usage in embedded systems. Despite of multiple proposals for improving the multi-core reliability, their use is not yet validated for critical tasks. This chapter describes a new fault-tolerance approach called NMR-MPar which is based on N-Modular Redundancy and M-Partitions to improve the reliability of applications running on these devices. The evaluation of the effectiveness of the NMR-MPar approach on two complementary benchmark applications running on the 28 nm CMOS MPPA-256 many-core processor has shown the possibility to consider this approach for mixed-criticality systems. Finally, this chapter analyses the overhead of the approach in terms of power consumption and energy.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    45
    References
    0
    Citations
    NaN
    KQI
    []