Multi-Objective Optimization of Channel Mapping for Fail-Operational Hybrid TDM NoCs

2019 Seventh International Symposium on Computing and Networking Workshops (CANDARW)(2019)

引用 2|浏览11
暂无评分
摘要
Despite their technical and economic advantages, multi-core processors are hesitantly adopted in safety-critical embedded application domains. A key issue is to provide hard real-time guarantees and fault-tolerance for critical applications as well as isolation between applications of different criticality on shared resources such as the interconnect. To address this, hybrid NoCs that combine configurable Time-Division-Multiplexing (TDM) for critical traffic with packet switching for Best-Effort (BE) traffic have been proposed. More recently, hybrid NoCs with protection switching schemes have been introduced to cope with fail-operational applications. Although they can handle failures happening in a NoC and guarantee real-time requirements by design, the problem of providing optimized mapping of critical tasks while considering the properties of protection switching schemes still remains. In this paper, we propose a multi-objective optimization methodology to tackle this matter. We first propose a model to map TDM channels onto a given NoC topology while considering protection switching schemes. The implemented objectives are defined so as to evaluate the behavior of the TDM mapping while still capturing its impact on the BE traffic, since a complete simulation considering BE would be time consuming (hundreds of thousands of NoC cycles with different injection rates). We then show how beneficial simultaneously optimizing several objectives is when using hybrid NoCs. The proposed methodology can indeed enable design space exploration in order to find feasible mappings for application specific scenarios and broad trade-off analyzes without having to run prohibitively long BE traffic simulations.
更多
查看译文
关键词
Hybrid NoC,TDM,Fail-Operational,Hard Real Time,Protection Switching,Fault-Tolerance,Channel Mapping,Multi-Objective Optimization
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要