Memory concepts for enabling adaptivity in distributed embedded systems
Establishing cost and resource efficient dependability through means of adaptivity in safety-critical distributed embedded systems is a strenuous endeavour, as the varying requirements on resilience, control and efficiency across domains prohibits a single solution to suit all needs. To assist the process of determining a safe and efficient system architecture with satisfactory precision, this work exemplifies the importance of differentiation by only addressing distributed embedded systems that perform multiple functions with alternating levels of criticality. Further, they do not require full fail-operational behaviour, thus allowing to sacrifice less important functions in the pursuit of preserving safety. Herein, a dynamic instantiation and graceful degradation strategy is developed to subsequently study its effect on cost when implemented in conjunction with execute-in-place (NOR-flash) or block-addressable (NAND-flash) memory concepts. Even though NOR-flash is generally considered to be a better candidate for such systems, this qualitative research produces evidence that NAND-flash memory concepts are likely to financially outperform traditional architectures when considering adaptivity.