Research On Key Technologies Of Large-Scale Graph Partition Stratigies

In recent years,due to the advent of the era of big data,the amount of global data presents an explosive growth.Graph data is of great value as a data structure that can describe complex relationships.Therefore,the study of graph calculation becomes more and more important.Faced with such a large scale of graph data,a series of research issues related to graph computing have become a hot topic in order to solve the problem of reasonable data storage and calculation.The problem of graph partition is the pre-processing step of graph calculation,which greatly affects the efficiency of graph calculation application.A good graph partitioning algorithm should satisfy the load balance of each partition and make the communication cost as small as possible.In order to achieve this goal,this topic first explores the influencing factors and degree of the effect of graph partitioning.From the three dimensions of graph data structure,data reading order,and the number of partition,a large number of experimental comparisons are carried out on different graph partitioning algorithms under the condition of control variables,and the rules are summarized.For the graph flow algorithm HDRF,sensitive to data flow sequence,lack of local information mining,as well as NE algorithm iteration times,low efficiency of partition.We put forward the local structure-based stream partitioning algorithm and the local structure-based vertex partitioning algorithm respectively.Experiments show that the two methods can effectively improve the partitioning efficiency and quality.