In listening to discussions about the library catalog, I am surprised not to hear more about how the type of library affects our assessment of how central the catalog is to library services or user behaviors.
For simplicity sake, think reductively of three categories of library material: bought, licensed and digital.
Bought materials (books, DVDs, CDs, ...) are typically managed within the integrated library system workflow, are cataloged and appear in the library catalog.
Licensed materials (e-journals, databases, ...) are typically managed within an emerging knowledge base/ERM/custom workflow, and appear to the user in a variety of databases, maybe consolidated through metasearch and resolver systems.
Digitized/digital materials (digitized collections, research and learning materials in repositories, ...) are typically managed within a repository environment, and appear to the user through a user interface to that environment.
Newer discovery layers may try to provide access across these three strands (as well as others), and sometimes data or services will be syndicated to other environments (e.g. Google Scholar, toolbar/widget, etc).
The systems to provide access to these three collection types probably account for the vast majority of access traffic to library collections.
However, volume of access breaks down differently across types of libraries. Digital is probably a minority in most. The catalog may get more traffic than access mechanisms for licensed materials in many public libraries, quite a bit more in many cases. And access mechanisms for licensed materials may get more traffic than the catalog in many academic environments, quite a bit more in many cases.
It would be interesting to synthesise recent research findings to quantify this ....