A Debian repository is a set of Debian packages organized in a special directory tree, together with some infrastructure (checksums, indices, signatures, desciption translations, ...), that allows debian users to easily update their system or install additional software.
Anatomy of a repository
The following description is mostly for people who browse a repository using a standard web browser and wonder what is where and how everything fits together. A more precise and technical description is in Format.
A debian repository contains several releases. Debian releases are named after characters from the "Toy Story" movies (wheezy, jessie, stretch, ...). The codenames have aliases, so called suites (stable, oldstable, testing, unstable) A release is divided in several components. In debian these are named main, contrib, and non-free and indicate the licensing terms of the software they contain. A release also has packages for various architectures amd64, i386, mips, powerpc, s390x, ...) as well as sources and architecture independent packages.
The root directory of a repository has a directory dists which in turn has a directory for each release and suite, the latter usually symlinks to the former, but the browser won't show you a difference. Each release subdirectory contains a cryptographically signed Release file and a directory for each component. Inside these are directories for the different architectures, named binary-<arch> and sources. And in these are files Packages that are text files containing the meta data of packages. Hmm, so where are the actual packages?
The packages themselves are below pool in the root directory of the repository. Below pool there are again directories for all the components, and in these are directories named 0, ..., 9, a, b, ... z, liba, ... , libz. And in these are directories named after the software package they contain, and these directories finally contain the actual packages, i.e the .deb files. The name is not necessarily the name of the package itself, the package bsdutils e.g resides in the pool/main/u/util-linux directory, it is the name of the source that the package is generated from. A single upstream source may generate several binary packages, and all of them will end up in the same subdirectory below pool. The additional single letter directories are just a trick to avoid having too many entries in a single directory which is what many systems traditionally have performance problems with.
In the leaf directories below pool there usually are several versions of a package, and the information what releases which version belongs to solely resides in the indices. This way, the same version of a package can belong to several releases but use disk space only once, and that without resorting to hardlinks or symlinks, so mirroring is rather simple and even works with systems that don't have these concepts.
The official debian package repository is mirrored all over the world.
The debian website has an official list of Debian mirrors
There also is a list of unofficial repositories containing various additional software
There is a team in charge of mirrors.
Working with repositories
There are two different ways of working with repositories: You can use it with the apt family of programs (apt, apt-get, apt-cache, aptitude) to browse or install packages, or you can set up a repository yourself and add remove or replaces packages in it.
Using a repository
A comprehensive guide showing best practices is in UseThirdParty
Set up and maintain a repository
There are various reasons for building a repository yourself. You might just have a few packages with local modifications that you want to make available to apt, you may want to run a local mirror with those packages used by several machines to save bandwidth, or you build packages yourself and want to test them prior to publication.