RDF Data Storage And Performance Evaluation Based On Redis

As an important storage framework in knowledge graph,RDF(Resource Description Framework)data storage is widely used in biological,medical,geographical and other fields.It has become one of the current research hotspots.Evolved from semantic web,knowledge graph has more promising applications in RDF data storage research.However,with the rapid increasing scale of Internet Data in the era of Big Data,how to store massive data effectively and efficiently has become a great challenge in the research of semantic web.Using traditional HTML and XML document are now difficult to handle the mass data storage and meet the processing requirements for semantic web.Many major technical companies and scientific research labs have invested the NoSQL database area.Therefore,storing RDF data by NoSQL is one of the main research directions in the field of semantic web.In view of the above challenges,this thesis studied Redis graph database,which is a representative key-value storage model based on memory storage.We also studied the traditional data storage model of RDF.The work contains the following three parts.Firstly,we focus on the challenge of handing mass data storage in today's semantic web domain,this thesis designed a RDF distributed data storage scheme based on Redis graph database.RDF triples are mapped into hash map structure in Redis by using the RDF graph structure.And the key-value model used ziplist which is an internal data structure,the same object of all predicates and values are packed into one hash map entry.Compared with other traditional storage models,this model is more concise and flexible with RDF storage,and can build distributed storage with Redis database.Secondly,in order to test our storage model's correctness and efficiency based on Redis database,this thesis designed a distributed performance test project based on Redis.The core idea is to distribute the resources evenly into the nodes,and use the SOA(Service-Oriented Architecture)framework for load control module to make every modules do their best in this test project.Finally,in order to make the performance evaluation works smoothly,we modified Redis database source code,and add some functions for RDF data,like add(),delete()and updata().So that the experimental operations are more convenient.