This chapter discusses various architectures for data integration among multiple database servers, which we refer to as peers. Specifically, we emphasize an expanded concept of data integration, termed Bidirectional Collaborative Data Management, wherein multiple peers exchange database updates via bidirectional updatable views. We provide an overview of several architectures by comparing key perspectives in bidirectional collaborative data management (Sect. 1), and delve into the details of our contributions, the architectures of Dejima (Sect. 2) and the BCDS Agent (Sect. 3). Dejima is a novel architecture that accomplishes database collaboration, where the connected peers can behave as if they share the same database through bidirectional transformation while selectively hiding parts of their own data. The BCDS Agent refers to the Dejima architecture and takes the compositional approach in developing the system which may be configured dynamically and changed by participants. We also discuss other related systems (Sect. 4). Additionally, we introduce two key techniques for data integration applications, recasting Schema Co-Existence (Sect. 5), and joinable table discovery (Sect. 6). The first technique supports different versions of schemas of the common database each used by its corresponding program along with data synchronization. The second technique finds joinable tables for data integration by leveraging embedding-based retrieval.