SELZON3D: seleção de zonas seguras para NOC-3D parcialmente conectada

Abstract

With the advancement of SoCs and the exponential growth in the complexity of MPSoCs, efficient and secure communication between multiple processing elements has become a critical challenge. Initially, architectures based on 2D-NoCs provided a scalable solution but proved to be limited as the number of components increased, leading to congestion and delays. To mitigate these issues, three-dimensional integration (3D-NoC) was adopted, promoting more efficient communication across layers. However, the implementation of vertical TSV connections in these systems faces challenges such as manufacturing defects and electromigration. Additionally, ensuring secure data traffic is essential, especially in sensitive applications such as IoT devices and healthcare systems. This work proposes a solution to the limitations related to the implementation of secure zones in a 3D-NoC context by applying the SELZON3D method, which selects secure zones in partially connected 3D-NoCs through the use of Tabu Search in the configuration of safe paths. This reduces the network’s average latency by optimizing packet delivery and ensuring data integrity. Through extensive simulations, varying traffic rates were injected until 3D-NoC saturation was reached. The results demonstrate up to a 41.07% improvement in performance compared to random assignments, ensuring reliable packet delivery with reduced latency. The integration of this method significantly enhances network stability and performance, highlighting the crucial role of secure zones in ensuring the reliable operation of MPSoCs.