Design And Implementation Of Graph Computing System Based On Ignite

The graph database is used to store and query graph data.With its efficient capabilities of processing for relationships,it is widely used in the financial industry,social networks,knowledge graphs,and network monitoring.In the era of big data,the high value of data is more reflected in the hidden information behind it.The graph database system GDM hopes to provide users with a comprehensive solution for graph data,but it is lack of analysis and processing for graph data.At present,graph computing systems for analyzing and processing graph data exist separately,and data migration before and after computation causes inconvenience to users.Therefore,a graph computing system is designed and implemented to be able to interface with GDM and provide analysis and processing capabilities for graph data.The implemented graph computing system is based on the Ignite open source framework,and can use its data storage and parallel computing capabilities to load and compute graph data.The system includes five functional modules.The data partitioning and loading module is used to convert the external original graph data,and divide the graph data into the Ignite cluster in the form of key-value data organization.The user interface module defines the vertex program interface that the user needs to implement,enables the system to interface with GDM,enables the user to access graph data with a common graph query language,and transfer and execute graph computation tasks.The computation flow control module is the core module of computation,and implements the GAS graph computing model.It adopts the vertex-centered traversal mode,calls the vertex program to complete multiple iterations of all vertices in the graph,and adds synchronization to the process to ensure the correct execution of graph computation tasks.and finally completes the update of graph data to obtain graph computation results.The message communication module provides support for the previous module,and is used to send and receive and cache update messages generated by the vertex during the graph computing process.The computation result merging and persistence module is used to further merge the computation results on the updated graph data obtained after the graph computation is completed,to acquire the final overall merge result,and write all results to the disk.Carry out functional test and performance test on the designed and implemented graph computing system.The analysis of the experimental results shows that the system can be connected to GDM,complete the partition loading of graph data,and correctly execute graph calculation tasks.The performance on the small-scale cluster meets the expected settings and the needs of analyzing and processing graph data.