云数据库中基于极大熵差分进化的负载评估算法

摘要

Due to the 2PC ( Two-Phase Commit) protocol, all transactions of DDBS ( Distributed DataBase System) will roll back if one of distributed nodes was overloaded, which makes the DDBS difficultly adapt to the big data' s environment, whose data are dynamic and random. In order to solve this multi-objective optimization problem, an evaluation method based on maximum entropy diffrential evulution was proposed to evaluate system' s load. First, the problem was reformulated as a non-smooth single objective optimization problem via evaluation function, and a smooth single objective optimization problem with parameter via the maximum entropy function, and then using the differential evolution algorithm to solve the converted problem. Experimental results show that the load evolution algorithm based on maximum entropy function method can evaluate the load in big data' s environment, avoid single-node bottlenecks, and improve system' s performance.%由于分布式关系型数据库基于两阶段提交协议的设计方式，使得系统如出现单节点瓶颈问题，数据库事务将全部回滚，从而造成巨大的系统开销，影响数据库在大数据环境下的应用。针对这一现状，提出一种基于极大熵差分进化的负载评估算法，利用评价函数法，将多目标优化问题转化为不可微的单目标优化问题，再利用极大熵函数，将不可微优化问题转化为一个带有参数的无约束优化问题，最后用差分进化算法对其进行求解，找出节点资源最优集，从而为过载节点的数据迁移提供了理论依据，也进一步实现了对云数据库的设计。实验结果表明，该算法能够提高系统的整体性能，有效避免单节点瓶颈问题。