What does it mean by disconnected data access architecture of ADO.Net?

ADO.Net introduces the concept of disconnected data architecture. In traditional data access components, you make a connection to the database system and then interact with it through SQL queries using the connection. The application stays connected to the DB system even when it is not using DB services. This commonly wastes the valuable and expensive database resource as most of the time applications only query and view the persistent data. ADO.Net solves this problem by managing a local buffer of persistent data called data set. Your application automatically connects to the database server when it needs to pass some query and then disconnects immediately after getting the result back and storing it in dataset. This design of ADO.Net is called disconnected data architecture and is very much similar to the connection less services of http over the internet. It should be noted that ADO.Net also provides the connection oriented traditional data access services.
Traditional Data Access Architecture

ADO.Net Disconnected Data Access Architecture

Another important aspect of the disconnected architecture is that it maintains the local repository of data in the dataset object. The dataset object stores the tables, their relationship and different constraints. The user performs operations like update, insert, delete to this dataset locally and finally the changed dataset is stored in actual database as a batch when needed. This greatly reduces the network traffic and results in the better performance.