Design Methodology for Responsive and Robust MIMO Control of Heterogeneous Multicores

Abstract

Heterogeneous multi-core processors (HMPs) are increasingly being deployed to meet the performance/power requirements of emerging workloads, demanding adaptive and coordinated resource management techniques to control the resulting complexity of such systems. While multiple input, multiple output (MIMO) control-theory has been applied to adaptively coordinate resources for single-core processors, the coordinated management of HMPs poses significant additional challenges for controllability, robustness, and efficiency due to their large space of cross-layer configuration parameters. This paper presents for the first time a methodology to design MIMO controllers for coordinated management of HMPs with formal guarantees. Our approach addresses the challenges of: system decomposition and identification; selection of suitable sensor and actuator granularity; and appropriate system modeling to make the system identifiable as well as controllable. We illustrate the practical applicability of our approach on an ARM big.LITTLE HMP platform running Linux, and demonstrate the efficiency and robustness of our method by designing MIMO-based resource managers.