Here are two diagrams taken from this document showing a Centralized vs a Distributed repository architecture.
Centralized Repository: Note that the "Document Repository" object is contained within the "Infrastructure Provider" along side the "Document Registry".
Distributed Repository : Here the "Document Repository" is now contained within each hospital's infrastructure while the "Document Registry" still remains within the centrally located "Infrastructure Provider".
So what does this all mean ? By "Central Repository", I mean a central, single database that contains all the documents in the enterprise. The "Distributed Repository" implies multiple databases. These databases can be centrally located however , within a single data center. This would still be a "Distributed Repository" since each repository sits in its own database.
What is the advantage of this model ? You can best understand this by comparing it to "Google". When we perform a search in Google, the search term goes out to an index of terms within Google's infrastructure. This "index" in analogous to the "Document Registry". The index or registry contains links (or addresses) to the locations of the actual data. The actual data may reside at different servers throughout the world. This "Distributed Repository" model works the same way, and just like Google, it can scale infinitely.
Because of the scalability of this "Distributed Model", this model can be the basis of a "SaaS architecture" where there is a need to be able to scale to meet the needs of a rapidly growing user base.
Once again, please check out this IBM document for more information.
Also, here are some other useful IBM documents: Clinical Imaging Repositories, Solution Architecture for Clinical Data Management