Safe Dynamic Reshaping of Reconfigurable MPSoC Embedded Systems for Self-Healing and Self-Adaption Purposes

Abstract

Multiprocessor system-on-chip (MPSoC) architectures are a huge challenge in embedded system design. This situation arises from the fact that available MPSoCs and related designs flows are not tailored to the specific needs of embedded systems. This work demonstrates how to provide self-healing properties in embedded MPSoC design. This is achieved by combining the features of a generic approach to create virtualizable MPSoCs out of off-the-shelf embedded processors with a methodology to derive system configurations, such as task-processor bindings, which are optimal in terms of safety and execution time. The virtualization properties enable a reshaping of the MPSoC at runtime. Thus, system configurations may be exchanged rapidly in a dynamic fashion. As a main result of this work, embedded multiprocessor systems are introduced, which dynamically adapt to changing operating conditions, possible module defects, and internal state changes. We demonstrate the figures of merit of such reconfigurable MPSoC embedded systems by means of a complex automotive application scenario mapped to an FPGA featuring a virtualizable array of eight soft-core processors.

A Decision-Making Method for Run-Time Structural Adaptation Of FPGA-Based SOCS to Variations in Workload, Power Budget, Die Temperature, and Hardware Resources

“…The commonly observed on-chip fault mitigation methods are discussed here. [98] proposes a mechanism to exchange software tasks between processors in MPSoC systems in order to increase the system reliability in presence of faults. Dynamic scheduling and mapping techniques are also employed for fault mitigation [98,99,100].…”

“…[98] proposes a mechanism to exchange software tasks between processors in MPSoC systems in order to increase the system reliability in presence of faults. Dynamic scheduling and mapping techniques are also employed for fault mitigation [98,99,100]. The authors of [101] present a variant-based method for mitigating permanent faults in FPGA-based heterogeneous systems; several variants of the same task are stored such that they occupy different reconfigurable regions.…”

A Decision-Making Method Providing Sustainability to FPGA-Based SoCs by Run-Time Structural Adaptation to Mode of Operation, Power Budget, and Die Temperature Variations

“…Methods such as power gating [21,22], Dynamic Voltage and Frequency Scaling (DVFS) and DFS [23][24][25][26], dynamic scheduling techniques [25,[27][28][29][30] and dynamic mapping techniques [25,[31][32][33], and dynamic task migration are used for power and/or thermal aware workload management. Methods such as Triple Modular Redundancy (TMR) and scrubbing [34][35][36], Built-in Self-Test (BIST) procedures [37], device reprogramming to avoid damaged regions [38], dynamic scheduling and mapping [39][40][41], run-time relocation based methods [42,43], and variant-based methods [44][45][46][47] are commonly used methods for fault management. Many systems deployed on FPGAs/ SoPCs make use of RTOS or similar management systems to adapt to different dynamic system parameters.…”

Run-Time Mitigation of Power Budget Variations and Hardware Faults by Structural Adaptation of FPGA-Based Multi-Modal SoPC

“…Dynamic scheduling techniques [27,[29][30][31][32] and dynamic mapping (or resource management) techniques [27,[33][34][35] are other methods used to achieve power and/or thermal aware workload management. Dynamic scheduling and mapping techniques are also employed for fault mitigation [15,36,37]. To ensure reliability in mission-critical systems like space-borne systems, the emphasis is on protection against and recovery from transient or permanent faults due to radiation effects.…”