Hierarchical Agent Based NoC with DVFS Techniques

Network-on-Chip (NoC) is a promising architecture in the many-core on-chip systems. A hierarchical agent based NoC architecture is proposed which enables the NoC to autonomously adjust itself, and provide maximum power efficiency, fault/variation tolerance and system flexibility. Agents are software or hardware components which monitor and control the system at different granularity. Via the joint efforts and interactions of agents at all levels of the architecture, system optimization can be achieved at the runtime. Agent hierarchy,mapping of agents on regular NoC platforms and the function partition among the agents are elaborately discussed. Runtime power management with various Dynamic Voltage and Frequency Scaling (DVFS) techniques is analyzed on the hierarchical agent based NoC platform. Conventional power monitoring techniques can be flexibly incorporated into the functions of specific levels of agents. Network condition is observed by the agents at the run time, and the power supplies are adjusted at different granularity accordingly. This paper describes the system architectures of two adaptive schemes, with efficient and feasible algorithms presented. Quantitative experiments demonstrate that they achieve superior power efficiency comparing with the basic architecture without any run time configuration, while different tradeoffs are applied dependent on the monitoring granularity.