SIL2LinuxMP strives to qualify a defined GNU/Linux subset for the use in safety-related systems by "assessment of non-compliant development". To demonstrate that the kernel has achieved suitable reliability and correctness properties basic metrics of such properties and their statistic analysis can be used as part of the argument. While IEC 61508 Ed 2 always pairs testing and analysis, we believe that for a high complexity system traditional testing is of relatively low effectiveness and analytical methods need to be the primary path.
In this talk we outline the used methods and give examples as well as key findings.
SIL2LinuxMP strives to qualify a defined GNU/Linux subset for the use in safety-related systems by "assessment of non-compliant development". To demonstrate that the kernel has achieved suitable reliability and correctness properties basic metrics of such properties and their statistic analysis can be used as part of the argument. Linux has a wealth of analytical tools built-in to it which allow to extract information on compliance, robustness of development, as well as basic metrics on complexity or correctness with respect to defined properties. While IEC 61508 Ed 2 always pairs testing and analysis, we believe that for a high complexity system traditional testing is of relatively low effectiveness and analytical methods need to be the primary path. To this ends we outline some approaches taken:
While much of the data points to robust and mature code there also are some areas where problems popped up. In this talk we outline the used methods and give examples as well as key findings. FLOSS development has reached a quite impressive maturity, to substantially go beyond we think it will need the use of quantitative process and code metrics - these results from SIL2LinuxMP may be a starting point.
Speakers: Nicholas Mc Guire