Byzantine Fault Tolerance. Ph.d., MIT, January 2001. Also as Technical Report MIT-LCS-TR-817. [CL+99] Xxxxxx Xxxxxx, Xxxxxxx Xxxxxx, et al. Practical byzantine fault tolerance. In OSDI, volume 99, pages 173–186, 1999. [CL02] Xxxxxx Xxxxxx and Xxxxxxx Xxxxxx. Practical byzantine fault tol- erance and proactive recovery. ACM Transactions on Computer Systems (TOCS), 20(4):398–461, 2002. [CML+06] Xxxxx Xxxxxxx, Xxxxxx Xxxxx, Xxxxxxx Xxxxxx, Xxxxxxx Xx- xxxxxxx, and Xxxxx Xxxxxx. Hq replication: A hybrid quorum protocol for byzantine fault tolerance. In Proceedings of the 7th symposium on Operating systems design and implementa- tion, pages 177–190. USENIX Association, 2006. [FLP85] Xxxxxxx X. Xxxxxxx, Xxxxx X. Xxxxx, and Xxxx Xxxxxxxx. Impos- sibility of Distributed Consensus with One Faulty Process. J. ACM, 32(2):374–382, 1985.
Byzantine Fault Tolerance. In Proceedings of the 3rd Symposium on Operating Sys. Design and Implem., pp. 173–186. [9] King, S. and Xxxxx, S, 2012. PPCoin: Peer-to-Peer Crypto- Currency with Proof-of-Stake. Online: xxxx://xxxxxxxx.xxx/assets/paper/peercoin-paper.pdf. [10] Xxxxxxxx, X., 2017. Review of Blockchain Consensus Mechanisms, Waves Platform. Online available: xxxxx://xxxx.xxxxxxxxxxxxx.xxx/review-of-blockchain- consensus-mechanisms-f575afae38f2 Accessed 2018. [11] Xxxxx, M., Xxxxxxx, X., and Xxxxxxx, X., 1980. Reaching Agreement in the Presence of Faults. Journal of the ACM 27, 228–234. [12] External Data Representation (XDR) Standard. Online: xxxxx://xxxxx.xxxx.xxx/html/rfc4506.html
Byzantine Fault Tolerance. Detecting a faulty process is easier – 2k+1 to detect k faults • Reaching agreement is harder – Need 3k+1 processes (2/3rd majority needed to eliminate the faulty processes) • Implications on real systems: – How many replicas? – Separating agreement from execution provides savings Reaching Agreement • If message delivery is unbounded, – No agreement can be reached even if one process fails – Slow process indistinguishable from a faulty one • BAR Fault Tolerance – Until now: nodes are byzantine or collaborative – New model: Byzantine, Altruistic and Rational – Rational nodes: report timeouts etc Compsci 677: Distributed and OS Lec. 18 13 Reliable One-One Communication • Issues were discussed in Lecture 3 – Use reliable transport protocols (TCP) or handle at the application layer • RPC semantics in the presence of failures • Possibilities – Client unable to locate server – Lost request messages t – Server crashes after receiving reques – Lost reply messages – Client crashes after sending request Reliable One-Many Communication •Reliable multicast – Lost messages => need to retransmit •Possibilities – ACK-based schemes • Sender can become bottleneck – NACK-based schemes Compsci 677: Distributed and OS Lec. 18 15 Broadcast Ordering • Broadcast (or multicast) order important for replication • FIFO broadcast: if a process sends m1 and then m2, all other processes receive m1 before m2 • Totally ordered: If a process receives m1 before m2 (regardless of sender), all processes receive m1 before m2 • Does not imply FIFO, all processes just agree on order • Causally ordered: if send(m1)->send (m2) => recv(m1)-> recv (m2) • State machine replication (SMR) • Broadcast requests to all replicas using totally ordered broadcast; replicas apply requests in order. Atomic Multicast •Atomic multicast: a guarantee that all process received the message or none at all – Replicated database example – Need to detect which updates have been missed by a faulty process •Problem: how to handle process crashes? •Solution: group view – Each message is uniquely associated with a group of processes • View of the process group when message was sent • All processes in the group should have the same view (and agree on it) Virtually Synchronous Multicast Compsci 677: Distributed and OS Lec. 18 17 Implementing Virtual Synchrony in Isis
Byzantine Fault Tolerance. In this subsection, we describe a key building block, namely the distributed consensus protocols in Byzantine [28] settings and highlight their importance to Blockchain systems. While there is a vast literature on Byzantine Fault Tolerance (BFT) consensus, we focus our discussion on PBFT and IBFT, which stand out for the following reasons. The former is the foundation for most of the research in this area, while the latter is implemented by many Blockchain systems today.
