The ability of the BigDataViewer (BDV), and of all the other software that is built around it, to display large images is achieved in great deal because of using chunks and resolution pyramids. Technically, this is achieved by storing the images, and their lower resolution copies, using the HDF5 file format. This brings a few advantages such as ability to use royalty-free and performant HDF5 library, data compression, and economical representation w.r.t. number of files in the file system by "baking" everything into the HDF5 container. But the compression and everything-inside principle makes it practically good only for create-once, read-only-forever type of access. With this setting, replacing a content, or adding another channel, cannot be done efficiently in general.
The N5 project, was designed not to use the everything-inside principle while using the chunks and resolution pyramids. And it offers, therefore, a smooth ability to replace content of chunks, or modify dataset "geometry" like adding another channel. User code controls N5 through its Java API, plus there is a few plugins in Fiji itself to load/store full images.
The DataStore is essentially following the N5 in the core functionality, it adds more on the client side by providing ImageJ2/Fiji plugins (and ImageJ macro commands consequently because any ImageJ2 plugin is callable from ImageJ macro). Additionally, it features client-server paradigm in a hope that it would make DataStore data accessible in more programming languages than just Java and even in a rather simple way.