BLOXY: Providing Transparent and Generic BFT-Based Ordering Services for Blockchains

2019 38th Symposium on Reliable Distributed Systems (SRDS)(2019)

引用 9|浏览50
暂无评分
摘要
With the wide-spread use of blockchain technology, Byzantine fault-tolerant (BFT) protocols are explored as a means to achieve consensus on which transactions should be processed next. BFT protocols are not a one-size-fits-all solution: they should be chosen according to the blockchain's use case, which can range from supply chain management to decentralised storage, requiring specialisation e.g. regarding throughput, latency, or level of decentralisation. Previously, consensus protocols were usually hardcoded into the blockchain infrastructure and could not be exchanged, therefore inhibiting flexible use of an otherwise generic blockchain infrastructure. Hyperledger Fabric claims to provide modular consensus and support for crash-fault and Byzantine fault tolerant protocols. However, integrating a BFT protocol has shown that Fabric's architecture is currently not well-suited for this fault model as it requires substantial changes and thereby breaks Fabric's modularity. This also has to be repeated for each integrated BFT protocol. In this paper, we present BLOXY, a blockchain-aware trusted proxy running on the replica that encapsulates all BFT client functionality. BLOXY enables transparent access to generic BFT frameworks and preserves Fabric's modularity even for the Byzantine fault model. It runs inside a trusted execution environment based on Intel's Software Guard Extensions. BLOXY offers blockchain-specific communication mechanisms as well as short-term block storage to handle crashes or disconnects to ensure that all nodes receive block updates. We implemented two BLOXY-based ordering services based on PBFT and the hybrid BFT protocol Hybster. Our evaluation shows that our approach increases the throughput of the ordering component by up to 71% compared to directly integrated BFT protocols.
更多
查看译文
关键词
Byzantine fault tolerance, Blockchain, Hyper-ledger Fabric
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要