The course aims to develop a thorough understanding of fault-tolerance in distributed systems. Due to their nature, distributed systems are inherently vulnerable to failures if not designed properly. At any time, a subset of the processes in a distributed system may fail by crashing or could be compromised and behave in a treacherous way (e.g., Byzantine failures). It is hence essential to design distributed systems and applications in such a way that they can adequately cope with failures. The lecture will present focus on how to deal with these issues.
By studying relevant methods and algorithms in details, the student will acquire a deep understanding of the issues at hand and the basic mechanisms to deal with such failures. Although the course will focus on the theory of such systems, it will also systematically draw links with practical applications, making it valuable to both theoreticians and practitioners.
Distributed algorithms, message-passing, synchrony models, agreement, replication, fault-tolerance, Byzantine agreement, blockchain, self-stabilization, blockchain, probabilistic algorithms
✔ 専門力 | 教養力 | コミュニケーション力 | 展開力(探究力又は設定力) | 展開力(実践力又は解決力) |
Typical classes will alternate between slide-based presentations, interactive discussions, class exercises. Active contribution to class discussions is strongly encouraged.
授業計画 | 課題 | |
---|---|---|
第1回 | Introduction, overview, reminder | Revision of basic concepts of distributed algorithms (models, synchrony, causality) |
第2回 | Models, faults, formalism, definition | 授業時に指示する. |
第3回 | Synchronous consensus | 授業時に指示する. |
第4回 | State-machine replication, replication techniques | 授業時に指示する. |
第5回 | Group membership, distributed transactions, atomic commit | 授業時に指示する. |
第6回 | Asynchronous Consensus, FLP impossibility proof | 授業時に指示する. |
第7回 | Unreliable failure detectors | 授業時に指示する. |
第8回 | Eventual leader election, Paxos | 授業時に指示する. |
第9回 | Randomized consensus | 授業時に指示する. |
第10回 | Byzantine consensus | 授業時に指示する. |
第11回 | Byzantine randomized consensus | 授業時に指示する. |
第12回 | Self-stabilization (definition, requirements, mutual exclusion, proof) | 授業時に指示する. |
第13回 | Self-stabilization (spanning-tree, distributed reset, composition, ...) | 授業時に指示する. |
第14回 | Distributed ledger and blockchain mechanisms | 授業時に指示する. |
第15回 | Q&A + final test | 授業時に指示する. |
Christian Cachin, Rachid Guerraoui, Luís Rodrigues, "Introduction to Reliable and Secure Distributed Programming," Springer, 2011. https://www.springer.com/jp/book/9783642152597
Reference Books:
1. Shlomi Dolev, "Self-Stabilization," MIT Press, 2000. https://mitpress.mit.edu/books/self-stabilization
2. Michel Raynal, "Communication and agreement abstractions for fault-tolerant asynchronous distributed systems," Morgan & Claypool, 2010. https://www.morganclaypool.com/doi/abs/10.2200/S00236ED1V01Y201004DCT002
3. Michel Raynal, "Fault-tolerant Agreement in Synchronous Message-passing Systems," Morgan & Claypool, 2010. https://www.morganclaypool.com/doi/abs/10.2200/S00294ED1V01Y201009DCT003
4. Wan Fokkink, "Distributed algorithms: an intuitive approach ," MIT Press, 2013.
5. Vijay K. Garg, "Elements of distributed computing," IEEE, 2002.
6. Gerard Tel, "Introduction to distributed algorithms (2nd ed.)," Cambridge Univ. Press, 2000.
7. Ajay Kshemkalyani, Mukesh Singhal, "Distributed computing: principles, algorithms, and systems," Cambridge Uni. Press, 2011.
Course materials:
Slide copies, additional article copies, ...distributed during lectures or made available for download from the course webpage.
Homework assignments and contribution to class discussion (30%), reports (30%), and examination (40%).
Examination will assess the understanding of basic concepts of fault-tolerant distributed algorithms (problems, algorithms, and methodology) and reasoning (correctness and complexity).
Required knowledge:
Prior to taking this course, the student must have previously acquired,
through lectures or self-study, background knowledge on basic concepts
of fault-free distributed algorithms such as taught in the following
courses:
- CSC.T438 Distributed algorithm; __or__
- MCS.T406 (CSC.T406) Distributed Systems
Related course:
In the field of fault-tolerant and dependable computing systems, this course is complementary with:
- CSC.T524 Dependable Computing