[12/09/23] Partner Event - Optimisation-Driven Scalable Byzantine Consensus and Its Applications About the event Abstract In the realm of distributed systems, Byzantine fault-tolerance (BFT) consensus stands as a foundational pillar, particularly vital in the context of technologies like blockchains. However, traditional implementations of BFT, such as classic PBFT and linear PBFT variants, grapple with issues of message communication complexity, thus limiting the scalability and performance of BFT algorithms, especially when confronted with the demands of large-scale systems and burgeoning peer numbers. To address these challenges, we introduce ParBFT, an innovative Byzantine consensus parallelism framework that amalgamates classic BFT protocols with a novel Bilevel Mixed-Integer Linear Programming (BL-MILP) optimization model. The overarching objective of ParBFT is twofold: to bolster scalability through parallel consensus and to fortify safety, ensuring steadfast total order consistency among correct replicas. A pivotal novelty lies in the integration of the BL-MILP model into ParBFT, empowering us to compute optimal numerical decisions for parallel committees, including the optimal committee count and peer allocation for each committee. This optimization not only advances consensus performance but also fortifies security. Additionally, we address the inherent intricacies of BFT state machine replication (SMR) and its adaptability to large-scale applications necessitating high performance while enhancing scalability. In response, we improve the ParBFT protocol with trusted execution environments (TEEs). Our proposed approach remarkably reduces the lower bound on the required number of peers from 3f+1 to just 2f+1 for better consensus performance while ensuring safety. The proposed protocols have been developed as a Java library which can be applied to one of the current blockchain platforms, i.e., R3 Corda. Extensive experiments and performance evaluations, conducted on a cloud platform testbed, demonstrate the superiority of our protocol. Speaker Bio Xiao Chen obtained his MSc and PhD degrees in computing science from Newcastle University. He held a research associate position at Arizona State University (US). Currently, he serves as a full-time Research Fellow under the Marie Sklodowska-Curie programme at the School of Informatics, University of Edinburgh (UK), and is set to take up a lecturer position at the University of Leicester (UK). His research primarily revolves around the theory of distributed systems and their practical applications, such as blockchains. Additionally, he specializes in stochastic and formal modelling techniques, as well as AI-driven optimization methodologies tailored for distributed systems. Registration details Registration: https://www.smartsurvey.co.uk/s/D8MKWE/ External link: https://meeting.huaweicloud.com/welink/j/97122598/4xDIvZLsbvSi5MlCUhQH0Hqg2s6jxtAI5 Meeting ID: 97122598 Passcode: 103446 Sep 12 2023 11.00 - 12.00 [12/09/23] Partner Event - Optimisation-Driven Scalable Byzantine Consensus and Its Applications Join Huawei and Xiao Chen for this tech talk, taking place on 12th of September. 4th floor Bayes Centre, Coffee House Register
[12/09/23] Partner Event - Optimisation-Driven Scalable Byzantine Consensus and Its Applications About the event Abstract In the realm of distributed systems, Byzantine fault-tolerance (BFT) consensus stands as a foundational pillar, particularly vital in the context of technologies like blockchains. However, traditional implementations of BFT, such as classic PBFT and linear PBFT variants, grapple with issues of message communication complexity, thus limiting the scalability and performance of BFT algorithms, especially when confronted with the demands of large-scale systems and burgeoning peer numbers. To address these challenges, we introduce ParBFT, an innovative Byzantine consensus parallelism framework that amalgamates classic BFT protocols with a novel Bilevel Mixed-Integer Linear Programming (BL-MILP) optimization model. The overarching objective of ParBFT is twofold: to bolster scalability through parallel consensus and to fortify safety, ensuring steadfast total order consistency among correct replicas. A pivotal novelty lies in the integration of the BL-MILP model into ParBFT, empowering us to compute optimal numerical decisions for parallel committees, including the optimal committee count and peer allocation for each committee. This optimization not only advances consensus performance but also fortifies security. Additionally, we address the inherent intricacies of BFT state machine replication (SMR) and its adaptability to large-scale applications necessitating high performance while enhancing scalability. In response, we improve the ParBFT protocol with trusted execution environments (TEEs). Our proposed approach remarkably reduces the lower bound on the required number of peers from 3f+1 to just 2f+1 for better consensus performance while ensuring safety. The proposed protocols have been developed as a Java library which can be applied to one of the current blockchain platforms, i.e., R3 Corda. Extensive experiments and performance evaluations, conducted on a cloud platform testbed, demonstrate the superiority of our protocol. Speaker Bio Xiao Chen obtained his MSc and PhD degrees in computing science from Newcastle University. He held a research associate position at Arizona State University (US). Currently, he serves as a full-time Research Fellow under the Marie Sklodowska-Curie programme at the School of Informatics, University of Edinburgh (UK), and is set to take up a lecturer position at the University of Leicester (UK). His research primarily revolves around the theory of distributed systems and their practical applications, such as blockchains. Additionally, he specializes in stochastic and formal modelling techniques, as well as AI-driven optimization methodologies tailored for distributed systems. Registration details Registration: https://www.smartsurvey.co.uk/s/D8MKWE/ External link: https://meeting.huaweicloud.com/welink/j/97122598/4xDIvZLsbvSi5MlCUhQH0Hqg2s6jxtAI5 Meeting ID: 97122598 Passcode: 103446 Sep 12 2023 11.00 - 12.00 [12/09/23] Partner Event - Optimisation-Driven Scalable Byzantine Consensus and Its Applications Join Huawei and Xiao Chen for this tech talk, taking place on 12th of September. 4th floor Bayes Centre, Coffee House Register
Sep 12 2023 11.00 - 12.00 [12/09/23] Partner Event - Optimisation-Driven Scalable Byzantine Consensus and Its Applications Join Huawei and Xiao Chen for this tech talk, taking place on 12th of September.
