基于谱聚类的BaaS资源负载均衡调度算法

BaaS resource load balancing scheduling algorithm based on spectral clustering

Fabric作为超级账本的核心项目,以其多通道的设计为用户提供更为隐私的交易空间,为了解决基于分布式架构下的多通道资源负载均衡问题,提出了基于NJW谱聚类的区块链即服务(BaaS)负载均衡调度算法SC-channel。该算法将平台子节点的数量作为划分类簇数量的依据,首先,基于通道采用peer之间的Jaccard系数构造相似矩阵;其次,计算拉普拉斯矩阵,求取前k个特征值和特征向量并将特征向量单位化;最后,用基于数量加权的k-means算法完成聚类。在Kubernetes平台上对这种方法进行验证,并与采用经典k-meansi的NJW算法默认调度算法下的资源负载均衡度做了比较分析。结果表明,采用基于谱聚类的BaaS资源负载均衡调度算法可提高资源利用的均衡程度,增强了平台的可用性与可靠性。

As one of the core frameworks of the Hyperledger,Fabric provides users with private transaction space with its multi-channel design.In order to solve the problem of multi-channel resource load balancing based on distributed architecture,a Blockchain as a Service(BaaS)load balancing scheduling algorithm SC-channel based on NJW spectral clustering was proposed.The proposed algorithm took the number of platform sub-nodes as the basis for classifying the number of clusters.Firstly,based on channel the Jaccard coefficient between peer was used to construct the similarity matrix.Secondly,the Laplacian matrix was calculated to obtain the first k eigenvalues and eigenvectors,and the eigenvectors were unitized.Finally,the feature clustering was done using the classical weight-based k-means algorithm.The proposed algorithm was validated on the Kubernetes platform and its resource balance degree was compared with those of the NJW algorithm using the classic k-means and the default scheduling algorithm.Theoretical analysis and experimental results show that the BaaS resource load balancing scheduling algorithm based on spectral clustering can improve the balance of resource utilization and enhance the usability and reliability of the platform.