Abstract: From time to time there have been different models of data integration to manage and analyze data. Also with the emergence of big data, the database community has proposed newer and better solutions to manage such disparate and large data. Also, the changes in the data storage models and massive data repositories on the web have encouraged the need for novel data integration models. In this article, we try to present a case of various trends in integrating data through different models. We present a brief overview of Federated Database Systems, Data Warehouse, Mediators and new proposed Polystore Systems with the evolution of architecture, query processing, distribution, automation and data models supported within those data integration models. The similarities and differences of these models are also presented. Also, the novelty of Polystore Systems with various examples is discussed. This article also highlights the importance of such system for integrating large scale heterogeneous data. Keywords: Data integration, Multi-database systems, Polystore systems. A Survey on the Evolution of Models of Data Integration Shashank Shrestha 1* and Subhash Bhalla 2 1 Department of Computer and Information Systems, University of Aizu, Aizu-Wakamatsu, Japan. Email: d8201104@u-aizu.ac.jp 2 Department of Computer and Information Systems, University of Aizu, Aizu-Wakamatsu, Japan. Email: bhalla@u-aizu.ac.jp *Corresponding Author I. IntroductIon There is a large volume of data on the web which are non- or semi-structured. The data are provided through diferent scientifc, organizational, institutional, and governmental repositories. These data need to be shared, exchanged and integrated. The data on the web usually have catalogs and libraries where there are links to access the data (URL or HTTP). These type of data are called Linked Data [1]. These data require integration to access information through multiple data stores. Over the past few decades, improvements in the variety of data have given rise to comprehensive research on the management of heterogeneous data. There are problems with the data sets available in various scientifc felds in combining data with diferent models. Integration of data is a method of combining multi-source data. Over all the underlying sources, it should have a single view. Generally, it has two types of architecture. Data Warehousing is a type of physical model where data from multiple sources are copied and stored in a warehouse [2]. Other type of architecture is the virtual architecture. The virtual architecture basically comprises of one of the following- Federated Database Systems (FDBS) [3], Mediation [4] and newly proposed Polystores [5]. The virtual architecture focuses mainly on the management of heterogeneous data with multiple sources/stores. The general term for the virtual architecture is Multi-Database Systems (MDBS). Since it proposes solution to manage heterogeneous data, MDBS can also be called Heterogeneous Database Systems. MDBS is designed to ofer the same functionality across various operating systems, data formats, and languages for queries. Database management systems (DBMS) running on heterogeneous computing platforms are usually managed by databases. Information sharing across various channels must also be provided by the MDBSs. Usually, databases are under diferent and autonomous supervision. In order to separate themselves from the distributed database architecture, the sources underlying the MDBS must have autonomy. There are various facets of autonomy that must be addressed by MDBS. They are classifed as autonomy of design, autonomy of communication, autonomy of implementation and autonomy of associations. The balance between autonomy and heterogeneity [3] is another aspect that must be considered. • Design Autonomy: Design Autonomy is the capacity, regardless of data, query language or conceptualization, to choose the design. • Communication Autonomy: The autonomy of communication is the general operation of DBMSs to communicate with or not with other DBMSs. • Execution Autonomy: Execution Autonomy is the permission of the DBMS component to manage the operations needed by local and external operations. International Journal of Knowledge Based Computer Systems 8 (1 & 2) June & December 2020, 11-16 http://www.publishingindia.com/ijkbcs/