Efficient multi-level fault simulation of HW/SW systems for structural faults

Abstract

In recent technology nodes, reliability is increasingly considered a part of the standard design flow to be taken into account at all levels of embedded systems design. While traditional fault simulation techniques based on low-level models at gate-and register transfer-level offer high accuracy, they are too inefficient to properly cope with the complexity of modern embedded systems. Moreover, they do not allow for early exploration of design alternatives when a detailed model of the whole system is not yet available, which is highly required to increase the efficiency and quality of the design flow. Multi-level models that combine the simulation efficiency of high abstraction models with the accuracy of low-level models are therefore essential to efficiently evaluate the impact of physical defects on the system. This paper proposes a methodology to efficiently implement concurrent multi-level fault simulation across gate-and transaction-level models in an integrated simulation environment. It leverages state-of-the-art techniques for efficient fault simulation of structural faults together with transaction-level modeling. This combination of different models allows to accurately evaluate the impact of faults on the entire hardware/software system while keeping the computational effort low. Moreover, since only selected portions of the system require low-level models, early exploration of different design alternatives is efficiently supported. Experimental results obtained from three case studies are presented to demonstrate the high accuracy of the proposed method when compared with a standard gate/RT mixed-level approach and the strong improvement of simulation time which is reduced by four orders of magnitude in average. CitationBaranowski R, Di Carlo S, Hatami N, et al. Efficient multi-level fault simulation of HW/SW systems for structural faults.

A fault-injection methodology for the system-level dependability analysis of multiprocessor embedded systems

“…When considering the aim of anticipating the dependability analysis at the early stages of the design flow, the usage of SystemC and TLM [10] is particularly appealing due to their support for the modeling of the system at different levels of abstraction, and in particular at the functional one. A large set of fault injection frameworks for SystemC/TLM specifications has been proposed in the literature (e.g., [9,[17][18][19]30]). However, most of these works mainly investigate fault injection strategies specific for the SystemC models [11,18,30] or define functional fault models for high-level specifications modeled with the TLM approach [31,17], neglecting the issues of the dependability analysis.…”

“…A large set of fault injection frameworks for SystemC/TLM specifications has been proposed in the literature (e.g., [9,[17][18][19]30]). However, most of these works mainly investigate fault injection strategies specific for the SystemC models [11,18,30] or define functional fault models for high-level specifications modeled with the TLM approach [31,17], neglecting the issues of the dependability analysis. Very few works proposed approaches for the system-level dependability evaluation in complex systems specified with SystemC/TLM [19,32,33,30].…”

“…However, most of these works mainly investigate fault injection strategies specific for the SystemC models [11,18,30] or define functional fault models for high-level specifications modeled with the TLM approach [31,17], neglecting the issues of the dependability analysis. Very few works proposed approaches for the system-level dependability evaluation in complex systems specified with SystemC/TLM [19,32,33,30]. For instance, the approach in [19] supports FMEA by specifying custom monitoring conditions on the primary output of the systems, thus taking into account the system functionality.…”

Multilevel Simulation of Nonfunctional Properties by Piecewise Evaluation

“…The concept of multilevel modeling has found widespread use for design validation [Paulin et al 2002], fault simulation [Baranowski et al 2011], as well as prediction of nonfunctional properties such as power [Brooks et al 2000]. This work does not compete with such specialized approaches, but rather provides the basic methodology for NFP prediction using multilevel models.…”