Distributed Consensus for Asynchronous Space Applications (CASA)

Distributed consensus is proposed to provide software-based modular redundancy for spaceflight applications, in order to overcome the risk of environmental effects - especially radiation - on digital avionics designs for spacecraft. Consensus for Asynchronous Space Applications (CASA) is an application developed by Astrobotic as a reusable, portable, extensible, and scalable solution for space systems requiring low overhead and low latency operations. Mission use cases for space systems that require performance and safety-critical constraints were chosen to develop CASA as a software application. CASA was developed for managing distributed consensus algorithms as a service to be used by other space applications and is implemented as an application in NASA's open-source flight software framework, Core Flight System (cFS). This mission-ready implementation leverages the hardware abstraction that cFS offers and enables a certain degree of hardware and platform agnosticism. Software-based distributed consensus, as implemented in CASA, is evaluated, here, as an alternative to commonly used hardware-based voter logic for modular redundancy in spaceflight. Radiation-tolerant designs for spaceflight applications often employ N-modular redundant processes to overcome radiation-induced faults and errors. These processes may be threads of code, combinational logic, entire applications, or board-level outputs. Redundant processes are joined as a voter domain behind common voter logic between their outputs. In spaceflight, voters are typically a radiation-hardened, hardware-based voter circuit. Voter logic takes input from N redundant process outputs, compares them, and outputs a single answer when a majority of the inputs are identical. While more complicated designs exist, single voters are a single point of failure in a system. In contrast, distributed consensus algorithms are robust against single-points-of-failure. These consensus algorithms provide a logical procedure for coordinating data and ensuring consistency between redundant processes, e.g., in a distributed computing cluster. Hardware redundancy carries a certain amount of overhead and constrains reusability. To overcome the constraints and complexity of hardware, CASA's distributed consensus approach focuses on a flexible software-based architecture for modular redundancy. This work investigates distributed consensus as an alternative to voters for fault-tolerant infrastructure in software for space systems with respect to dependability, latency, and resiliency. This paper presents the background for distributed consensus, its application for space systems, use cases for CASA in real space missions, the testing methodology, discussion of this work's preliminary test results within a context of overhead and reconfiguration costs, and direction for future work.