Efficiency analysis of ontology servers

One of the main current trends in the ICT sector is the widespread use of intelligent devices. We can witness the expansion of the IoT architecture where smart devices can interact with each other and they can operate as intelligent agents to control our environment. Usually, these devices have a local knowledge base providing information for the concrete actions. In order to implement a higher level of intelligence, the devices should interact with humans in a more general environment. The main representation form of common sense knowledge in intelligent applications is ontology. Consequently, a widescope and detailed knowledge base that can cover general requirements should be stored in a general ontology server. This means that ontology databases should be able to manage huge amount of data in an efficient way. Although the concept of computerized ontology management is almost 10 years old (Gruber introduced this term in 2009), currently we have experiences only with smaller and separate ontologies. We can expect though that the future development trends in IoT will require the implementation of new kinds of large scale ontology servers, too. This paper presents first the main requirements on a general ontology server focusing on the demands in an IoT architecture. The main goal of the paper is then to compare the efficiency of three implementation alternatives: native RDF triplet storage, relational database storage and graph database storage.