PhD position: Designing Fault-Tolerant Distributed Systems for Nanosat Constellations/Swarm of drones

Keywords: orbital edge computing, distributed algorithms, cyber-physical systems, replication, distributed and resilient computing, fault tolerance, intermittent computing, nanosats’ subsystems.

Context
——————————

The recent deployment of an increasing number of nanosatellites in low-earth orbit (LEO) presents new opportunities for space applications [DKL + 17, DL20]. Built atop small-sized yet powerful blocks,a.k.a. CubeSats or simply nanosats, nanosatellite constellations emerge as promising platforms for massive sensing and large-scale distributed computing. Indeed, they represent a cheaper, competitive alternative for traditional satellite systems for a wide range of application domains such as earth observation and defence.
However, the design of distributed, intelligent systems based on nanosats is particularly challenging: nanosats have more stringent physical limitations with respect to processing/networking capability, energy supply, and connectivity among nanosats. Moreover, the use of cheaper components and subsystems might expose the emerging nanosat applications to performance degradation or complex failures. Therefore, novel resilient distributed applications and protocols should be designed and evaluated to make efficient and reliable use of the resources of nanosats at the orbital edge. The proposed doctoral project aims to enable a first-of-its-kind orbital edge computing subsystems with nanosats and to design novel techniques to support reliable and efficient data processing for emerging sensing application like earth observation with the proposed orbital edge computing platform. In order to achieve this challenging goal, we will conduct interdisciplinary, collaborative research to answer the following questions:
+ How to enable distributed computing on a nanosat?
We will survey the design and implementation of state-of-the-art building blocks including suitable communication protocols and specific subsystems interfaces and abstractions for computing on nanosats.

+ How to build a resilient computing system with a set of nanosats?
We will investigate distributed systems problems and propose specific solutions for dynamic reconfiguration mechanisms, consensus algorithms, and data replication schemes on nanosats systems. For that, we will take into account the ongoing research on related topics at CNES, including clock synchronization.

Proposed research
——————————

This doctoral research project aims to address the above scientific challenges as follows:

+ Leveraging non-expensive, failure-prone nanosats’ components. We will explore the design space and the performance evaluation of distributed system on nanosats constellations. Based on an existing, representative hardware platform proposed by the CNES, we aim to conduct a systematic study on how different choices of distributed systems primitives and designs affect the performance of key services, such as special-purpose sensing and distributed processing application. To this end, we will execute specific benchmarks to identify design opportunities, to assess the impact of different failures and to better understand the eventual trade-offs for distributed computing on nanosats.

+ Resilient edge computing with constellations of nanosats. An interesting solution for processing large amount of sensing data is to build a distributed computing system with a set of nanosats. So that we will re-examine many assumptions in traditional distributed systems in the presence of processing and interconnectivity limitations of nanosats. In particular, we aim to design novel resilient applications and protocols for fault-tolerant distributed services, e.g., dynamic reconfiguration mechanisms, consensus protocols, and replication schemes. Based on these fundamental services, we will enable intelligent, distributed computing on nanosats constellations.

Currently, the availability and resilience of traditional, cloud-based distributed system are commonly guaranteed by a replication protocol based on replicated state machine (RSM). Such a protocol implements a consensus algorithm to enable strong consistency, like Fast Paxos [Lam06] and Raft [OO14]. Strongly consistent replication is key to efficient implementation of critical distributed systems’ building blocks, like distributed lock manager, reliable configuration or transactional key-value store. To our knowledge though, such protocols have never been designed and extensively evaluated on nanosatellite constellations. Finally, it is important to highlight that this doctoral project is a joint work with our colleagues in University of Adelaide in Australia, therefore the doctoral student will spend at least one year working in Australia. In addition, the doctoral candidate will have a joint doctoral degree by the end of the project.

References
————————–
[DKL + 17] K. Devaraj et al. Dove high speed downlink system. 2017.
[DL20] B. Denby and B. Lucia. Orbital edge computing: Nanosatellite constellations as a new class of computer system. In the ASPLOS, 2020.
[Lam06] L. Lamport. Fast paxos. Distributed Computing, 2006.
[OO14] D. Ongaro and . Ousterhout. In search of an understandable consensus algorithm. In the ATC, 2014.