Ontology as the base of semantic web is used in many applications. Different ontologies in the same domain lead some heterogeneities and ontology matching systems are developed for resolved them. Heterogeneities have arisen owing to the fact that these ontologies have been created by various people through diverse methods. Nowadays, using large-scale ontologies in some applications such as medical fields seems inevitable. By using large-scale ontologies, some problems like the shortage of memory consumption and long duration of execution appeared in ontology matching systems. In this paper, large-scale ontology matching systems are studied and proposed a general architecture for them. Then large-scale ontology matching systems classified based on the partitioning large ontologies into several sub-ontologies, as known as the modularization, decomposition, summarization, clustering, and divide and conquer categories. This new classification will be useful for future research works in this field. In order to find out the efficiency of the ontology matching systems the results of OAEI (Ontology Alignment Evaluation Initiative) for the period 2011 to 2015 are compared. In spite of great progress, increasing accuracy is required in some section such as conference and benchmark sections.