Doctoral research project

Person in charge of the project:
CATTHOOR FRANCKY, member of research team Associated Section of ESAT - INSYS, Integrated Systems
Proactive Hard Constraint Management in Cost-conscious Nonlinear DynamicComputing Systems
Project summary:
With the relentless scaling of CMOS technologies, variability and reliability are becoming a great threat to the design of reliable complex digital systems-on-chip (SoCs). Increasing electrical and thermal stresses,increasing importance of structural uncertainties, and lack of thoroughunderstanding of degradation mechanisms and their interactions in new materials and ultra-scaled structures are rendering technologists unable to meet failure rate targets of real-time embedded SoCs. From the chip designers' perspective, variability and reliability manifest as uncertainties in electrical parameters of devices. In sub-45nm era these device-level parametric uncertainties would be way too high to handle with prevailing worst-case design techniques without incurring significant penaltyin terms of area/delay/energy. State-of-the-art defect-tolerant and better-than-worst-case design techniques either can not deal with the type and amount of expected uncertainties or can not guarantee hard application deadlines in the face of time-varying or dynamic uncertainties. We foresee the delegation of reliability qualification not just to designers but even to a run-time manager of the SoC itself with appropriate degradation models from technologists and designers. This research projects aims to understand and classify the control mechanisms behind such run-time managers which need to optimize the system cost amidst conflicting hard constraints such as timing, lifetime, thermal safety, and quality of service or user experience. We formulated the basis of low-overhead trulyproactive controllers which can maximally exploit the system correlations to achieve near-optimal run-time control. Dynamic bounding, turbo modes, and proactive system conditioning are the key proposed concepts. We have demonstrated proactive control in the context of video decoding system by systematically applying these concepts. We are in the process of applying these concepts to more diverse systems to illustrate the generality of the proposed concepts.
ph.D student :
Faculty of Engineering Science
Doctoral Programme in Engineering Science (Leuven)