【作者】 许亮；

【导师】 郭兵；

【作者基本信息】 四川大学 ， 计算机科学与技术， 2021， 硕士

【摘要】 区块链作为一种去中心化的新型分布式计算范式,具有多方创造信任、安全自治、不依赖可信第三方等特性,得到了越来越广泛的关注和应用,从最初的数字货币应用场景,逐渐拓展到了金融、贸易、征信、版权保护等诸多行业领域。联盟链相较于公有链拥有更高的交易性能和良好的监管特性,广泛受到各组织机构的青睐,其应用场景也在各行业领域得到了极大的拓展。联盟链在发展和应用过程中也遇到了各种问题与挑战,特别是联盟链的可扩展性问题成为其进一步发展的瓶颈。联盟链的可扩展性问题主要包括三个方面:一是性能可扩展问题,虽然联盟链的交易性能明显高于比特币等公有链,但依然不能满足一些高并发高速率交易应用场景的需求。二是存储可扩展问题,由于全网节点都存储和维护相同的账本数据,无法通过增加节点的方式降低存储开销,而随着区块数据的不断增长,节点的存储负担将越来越重。三是节点动态可扩展问题,由于联盟链具有准入设计和监管机制,而基于拜占庭容错共识算法的联盟链系统中,节点不能动态的加入和退出共识网络,从而导致节点的动态扩展性受到制约。针对以上问题,本文重点对联盟链的可扩展关键技术进行研究,主要工作成果如下:一是针对联盟链交易性能仍然不够高,不能满足高并发高速率应用场景需求的问题,本文从分布式计算的角度出发研究联盟链性能扩展方案,设计了可扩展联盟链架构,提出了联盟链分布式计算模型。通过对联盟链网络进行功能分区,对系统节点进行更细致的逻辑分类,实现计算与存储分离。把交易的执行与验证环节剥离出来进行更细粒度的分解,实现交易执行与验证解耦合,通过节点动态分组并行计算的方式,提高交易执行的并行度,从而提升联盟链系统整体交易性能。二是针对联盟链存储可扩展问题,研究提出基于RS纠删码的分组负载均衡存储扩展方案。方案对联盟链多副本全量冗余存储模式进行调整优化,通过在系统中设置轻节点,将区块数据文件切分后使用纠删码进行编码,每个轻节点仅存储部分编码分块,有效减少了各节点的存储开销。方案采用延迟编码策略,以适当的冗余存储来提高系统存储容错率,并采取分组存储模式,通过设计算法来确保系统整体计算和存储负载均衡。三是针对联盟链节点动态可扩展性差,在拜占庭容错共识机制下节点不能动态加入和退出的问题,研究提出了基于推荐机制的节点动态扩展方案。方案采用推荐方式实现节点动态加入网络,并在PBFT共识机制上添加节点动态管理协议,在不需要停止和重启系统网络的前提下,通过共识流程动态更改全网共识状态参数,实现节点的动态加入和退出,以提高联盟链节点动态扩展性。最后,基于本文研究工作对超级账本Fabric架构进行改进,设计实现了数字版权交易系统,并基于版权交易系统对本文研究内容进行实验测试。实验结果表明,与Fabric原架构相比,改进的架构在交易性能、存储性能和节点动态扩展性等方面均有明显提升。更多还原

【Abstract】 Blockchain as a new decentralized distributed computing paradigm.On account of the properties of multi-party formulation of trust,security,autonomy,and sovereignty without third parties,it has earned rising attention and applications.Initially,blockchain was only applied in the digital currency industry.With its evolution,the adoption of blockchain has progressively expanded to various industries such as finance,trade,credit investigation,and copyright protection,etc.Compared with the public blockchain,the consortium blockchain has more eminent transaction performance and is more manageable to be supervised.Accordingly,the consortium blockchain is promoted by numerous industry organizations.The utilization scenarios of it have been exceedingly expanded in various industry fields.The consortium blockchain has encountered multiple difficulties and trials in the advancement.Among these hurdles,the bottleneck predicament that most affects its further growth is the scalability of the consortium blockchain.The scalability issue comprises three aspects.The first one is performance scalability.Specifically,though the transaction performance of the consortium blockchain is significantly better than that of public blockchain such as Bitcoin,it still cannot satisfy some application scenarios which demand high concurrency and high-rate performances.The second aspect is the storage scalability issue.For all nodes in the entire network property and manage the same block data,it is impracticable to lessen storage overhead by appending nodes.With the continuous germination of block data,the storage pressure of nodes will inevitably become much more onerous.The third aspect is the node dynamic scalability issue.The consortium blockchain has mechanisms of admittance and administration.Yet,in the Byzantine fault-tolerant consensus algorithm,on which the alliance chain system is based,nodes are not able to dynamically enter or exit the consensus network.This assuredly restricted the dynamic scalability of the node scale.Consequently,this paper concentrates on the key technologies of the scalability of the consortium blockchain.The main work results of this paper are as follows:Firstly,this paper attempts to solve the low transaction performance difficulty to fulfill the requirements of several high-concurrency and high-rate application scenarios.From the aspect of distributed computing,this paper analyzes the performance expansion of the consortium blockchain,devises a scalable alliance chain construction,and proposes a distributed parallel computing model for the consortium blockchain.Through the functional partition for the consortium blockchain network and broader specific logical classification of the system nodes,the research has accomplished the detachment of computing and storage.Separating the transaction execution and verification could present a more fine-grained decomposition,which can be used to achieve the decoupling of transaction execution and verification.Manipulating dynamic grouping and parallel computing of nodes,the study promotes the parallelism of transaction execution,thereby promoting the overall transaction performance of the consortium blockchain system.Secondly,in terms of the storage scalability concern,the research introduced a package load balancing storage expansion model that stands on RS erasure codes.This model fixes and optimizes the full redundant storage mode of compound copies of the ledger data in the alliance chain.By arranging light nodes in the system,the model applies erasure codes to encode the block data files after segmentation.In this way,per light node could only be stored as part of the coded blocks.This efficiently lessens the storage burden of individual nodes and presents proper redundancy to recover the storage fault tolerance of the system.This scheme adopts the group storage mode and assures the overall storage load balance of the system through the device of the algorithm.Thirdly,the paper contributes a dynamic node enlargement scheme based on the recommendation mechanism after regarding the feeble dynamic scalability and the inability of consortium blockchain nodes to dynamically enter and exit the consensus network under the Byzantine fault-tolerant consensus mechanism.This purpose accomplishes nodes to dynamically join the network,and appends a node dynamic management protocol to the PBFT consensus mechanism.Without the call to pause and restart the system network,node dynamics can dynamically adjust the consensus state parameters of the entire network to succeed in entering and exiting the consensus network.This purpose could hugely enhance the dynamic scalability of consortium blockchain nodes.Sequentially,this paper promotes the Hyperledger Fabric architecture,composes and executes a digital copyright trading prototype system.Based on the copyright trading system,the research conducted laboratory tests on the research content.Experimental results confirm that compared with the primary Fabric architecture,the refined architecture has notable advances in transaction performance,storage performance,and node dynamic scalability.更多还原