Efficient storage and querying of semantic data on a distributed system

One of the most important features of the Internet is that it lets information to be connected to each other. In this way, regardless of the publisher of it, contents related to each other, is accessible in the form of a network. However, the abundant number and diversity of the data on Internet, makes it difficult for users to quickly access and process the information they need. In the case that, computers can understand and interpret the data, preparation of information by software that users need, would be much faster. The greatest obstacle for computers to perform these functions is that very large portion of the data available on the Internet is non-structural data which was written in human natural language. To eliminate this obstacle, in addition to the usage of data just written in natural language that can be understood by the people, usage of semantic data which can be understood by computers was recommended. Mentioned new data will be added as a new layer to Internet and will enable computers to understand, interpret and use the new semantic network. Semantic data can be published by using languages such as the Resource Description Framework (RDF), Web Ontology Language (OWL) and the Extensible Markup Language (XML). This data is stored and queried in a structural form in customized database systems which are called Triple-Store. However, a large part of these solutions, the design of which are running on a single computer, does not satisfy the need for data to be scalable. Another group of solutions, use relational databases for the purpose of RDF Storage. In this case, when the amount of data is very large, queries take very long time as RDBMS's are not optimized to store Graph Data. In addition, relational databases can not provide inference capability which is carried out using OWL as it is performed by triple-stores. In this thesis work, a distributed infrastructure that can be used to store and query semantic data has been proposed and compared with a single Triple-Store running on a single computer. The proposed system solution has been implemented by programming it and tested with LUBM test queries on a set of artificial data which had been generated one of the most commonly used data set generator, called LUBM Data Generator. As the proposed system solution is a distributed one, query processes run in parallel on smaller data sets. The effect of this situation on the time interval for answering queries, have been tested with clusters including different number of units and results are shown in a comparison chart.