Optimization method and device of distributed graph database, electronic equipment and storage medium

Abstract

The invention discloses an optimization method and device of a distributed graph database, electronic equipment and a storage medium. All vertexes connected with edges of the same type and related information are stored on a single node of a distributed system; information contained in the incidence relation retrieval at least comprises information of all types of edges contained in the whole graph, information of unique nodes where the edges of each type are located, the total number of the edges of each type and related information of all data views to which the edges of each type belong. And the large-scale graph data is ensured to be stored on a plurality of system nodes in an optimized form by realizing association relationship retrieval through design. The retrieval comprises the number of vertexes connected with each type of edges, and a distributed graph database can be realized based on a load balancing algorithm. Meanwhile, by retrieving the information of the edges, parallel operation and processing for the distributed graph database are supported, and data consistency can be guaranteed.

Description

technical field [0001] The invention relates to the technical field of graph databases, in particular to an optimization method, device, electronic device and storage medium for a distributed graph database. Background technique [0002] Graph data (Graph Data) includes vertices and their attribute data that form a graph, edges and their attribute data, and there must be an edge between vertices with an associated relationship. Since graph data is different from traditional two-dimensional table structured data, when large-scale graph data is stored and processed in a distributed manner, a processing mode similar to structured data cannot be used (that is, decomposing a large form into multiple The reason is that if a large graph is directly divided into several subgraphs for storage, it is still necessary to design an appropriate algorithm to record and update the relationship between two vertices across subgraphs, which will affect the entire distribution. The operation e...

AI Technical Summary

Problems solved by technology

Since graph data is different from traditional two-dimensional structured data, when performing distributed storage and processing of large-scale graph data, a processing mode similar to structured data cannot be adopted (that is, a large table is decomposed into multiple sub-table for storage), the reason is that if a large graph is directly divided into several sub-graphs for storage, it is still necessary to design an appropriate algorithm to record and update the association between two vertices across sub-graphs, which will affect the entire distribution. The operating efficiency of the graph database, especially when performing basic operations such as graph data traversal, will be very complicated

[0003] Moreover, due to the continuous update and change of data, the topological structure of the graph is also changing, which will lead to unbalanced node load in distributed storage, that is to say, some subgraphs may become larger and larger and cannot be completely stored in one On the node, only subgraphs can be re-split to achieve load balancing, which will also become a complex task for distributed graph databases

Images