AXI-REALM: A Lightweight and Modular Interconnect Extension for Traffic Regulation and Monitoring of Heterogeneous Real-Time SoCs

Thomas Benz,Alessandro Ottaviano,Robert Balas,Angelo Garofalo,Francesco Restuccia,Alessandro Biondi,Luca Benini
2023-11-16
Abstract:The increasing demand for heterogeneous functionality in the automotive industry and the evolution of chip manufacturing processes have led to the transition from federated to integrated critical real-time embedded systems (CRTESs). This leads to higher integration challenges of conventional timing predictability techniques due to access contention on shared resources, which can be resolved by providing system-level observability and controllability in hardware. We focus on the interconnect as a shared resource and propose AXI-REALM, a lightweight, modular, and technology-independent real-time extension to industry-standard AXI4 interconnects, available open-source. AXI-REALM uses a credit-based mechanism to distribute and control the bandwidth in a multi-subordinate system on periodic time windows, proactively prevents denial of service from malicious actors in the system, and tracks each manager's access and interference statistics for optimal budget and period selection. We provide detailed performance and implementation cost assessment in a 12nm node and an end-to-end functional case study implementing AXI-REALM into an open-source Linux-capable RISC-V SoC. In a system with a general-purpose core and a hardware accelerator's DMA engine causing interference on the interconnect, AXI-REALM achieves fair bandwidth distribution among managers, allowing the core to recover 68.2 % of its performance compared to the case without contention. Moreover, near-ideal performance (above 95 %) can be achieved by distributing the available bandwidth in favor of the core, improving the worst-case memory access latency from 264 to below eight cycles. Our approach minimizes buffering compared to other solutions and introduces only 2.45 % area overhead compared to the original SoC.
Hardware Architecture
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the resource - sharing conflict problem arising from the co - existence of multiple processors and hardware accelerators on the same platform in modern heterogeneous real - time systems, especially in the automotive industry. Specifically, the paper focuses on how to provide system observability and controllability in these systems to ensure time predictability and the accuracy of compositional analysis, thereby avoiding critical tasks missing deadlines due to resource competition. To meet this challenge, the authors propose a lightweight, modular, and technology - independent real - time extension scheme named AXI - REALM, which is based on the industry - standard AXI4 interconnect design and aims to solve these problems through traffic regulation and monitoring at the hardware level. The main functions of AXI - REALM include: 1. **Hardware - driven traffic controllability**: Through the number of slave regions configurable per manager, the address range, budget, and reservation period configurable at runtime, AXI - REALM can isolate malicious attackers and allow the configuration registers of privileged access units. 2. **Hardware - driven traffic observability**: The Monitoring and Regulation (M&R) unit tracks the access and interference statistics of each manager, such as transaction latency, bandwidth, and interference with other managers, and performs bandwidth throttling according to the budget of each manager. 3. **Implementation - independent design**: AXI - REALM is integrated into an open - source Linux - compatible RISC - V SoC and characterized at the 12 - nanometer process node. Compared with the original design, AXI - REALM only incurs an area overhead of 2.45%. Through these features, AXI - REALM can allocate and control bandwidth in multi - subordinate systems according to the cycle - time window, actively prevent denial - of - service attacks from malicious actors in the system, and track the access and interference statistics of each manager to optimize the selection of budgets and cycles. This not only improves the performance of the system but also ensures the time predictability and reliability of critical tasks.