基于非冗余分配的分布式数据分配策略

Distributed data allocation policies based on non-redundant allocations

随着数据库技术的广泛应用和全面发展,如何提高系统性能已成为一个重要的研究课题。在提高系统性能这一过程中,降低各个站点上发生的各种事务的总代价,以减小成本和消耗,是必须进行深入探索的研究方向。对分布式数据库系统来说,事务的主要代价发生在站点间的数据传输过程中。数据分配的应有之义是生成一个合适的分配方案,并依此将数据段分配到每个站点,以最小化每个事务产生的数据传输量。因此,对数据分配策略的选用,将从进程上深刻影响分布式数据库的性能。从传统角度来看,将x个数据段分配给y个站点是一个NP完全问题,使用常规的穷举法必然会花费大量时间,造成资源上的浪费。因此,有必要根据数据库、应用程序、站点和网络等特定统计信息以及成本公式,制定并使用适当的分配策略。文章通过成本优化方法研究了数据分配问题,通过多组实验对非冗余式的分布式数据分配策略进行了证实。经过对比验证,该分配策略在每个实验中相比于已有的分配策略或其他可能的思路,都有较明显的优势。

With the wide application and all-round development of database technology,how to improve system performance has become an important research topic.In the process of improving system performance,reducing the total cost of various transactions on each site to reduce costs and consumption is a research direction that requires in-depth exploration.For distributed database systems,the main cost of transactions occurs in the process of data transmission between sites.The proper meaning of data allocation is to generate an appropriate allocation scheme and allocate data segments to each site to minimize the amount of data transmission generated by each transaction.Therefore,the choice of data allocation strategy will deeply affect the performance of distributed databases in terms of the process.From a traditional perspective,assigning x data segments to y sites is a complete NP problem[2].Using the conventional exhaustion method will inevitably take a lot of time and waste resources Therefore,it is necessary to develop and use appropriate allocation strategies based on specific statistical information such as databases,applications,sites and networks,as well as cost formulas.In this paper,the data allocation problem is studied through the cost optimization method,and the non-redundant distributed data allocation strategy is confirmed through multiple groups of experiments.After comparison and verification,the allocation strategy has obvious advantages over existing allocation strategies or other possible ideas in each experiment.