System-Level Dynamic Variation Margining in Presence of Monitoring and Actuation

Adaptive system has become a popular approach to mitigate dynamic variations with the help of monitoring and actuation. However, design margin is still required due to imperfections in the entire adaptation process, including inaccuracy and latency for both monitors and actuators. In this letter, we study the system-level margining problem in presence of monitoring and actuation. We first analyze the margining strategies for different types of dynamic variation sources. Case studies are performed with different monitor and actuator types for temperature variations. Our experiment results show that inappropriate design and selection of monitor and actuator can result in up to 2.8X more system margin. Based on the results, some design guidelines are given for system design optimization.