Dynamic Job Scheduling using the Least Utilized Cores for Enhanced Performance and Thermal Management in WNoC

Multicore architecture featuring wireless network-on-chip (WNoC) has emerged as a promising solution to meet the increasing demand for performance and power efficiency. However, the frequent utilization of cores to accomplish computational and communication objectives often leads to significant heat generation within multicore processors. This excessive heat poses the risk of overheating and potential malfunction of the processor chip. Multicore WNoC systems, whether with uniform or non-uniform subnets, often suffer from inefficient core utilization, which can result in decreased performance, localized overheating, and reduced processor chip lifespan. In this study, we propose a dynamic job scheduling methodology combined with the least used core selection to enhance performance and achieve even heat distribution across multicore WNoC chips. The proposed method employs the shortest remaining job first (SRJF) algorithm to select jobs and the least recently used (LRU) algorithm to select cores for ensuring even time distribution for equal heat dissipation while maximizing resource utilization. We simulate a 49-core 7x7 WNoC utilizing 1-hop (where a core needs only one hop to reach the nearest wireless router), 2-hop, and 3-hop organizations. Simulation programs are executed using a representative workload consisting of 92 jobs to evaluate the effectiveness of the proposed method. Our simulation results demonstrate how the communication latency and power consumption are impacted by the number of wireless routers. Also, it is observed that the proposed method effectively achieves balanced heat distribution across the WNoC. The 2-hop WNoC emerges as the optimal WNoC, achieving a 57% reduction in wireless routers compared to the 1-hop architecture, along with latency reductions of 9% and 14%, and power consumption reductions of 18% and 25% compared to the 3-hop and non-uniform configurations, respectively.