Now that we have completed our humble attempt at a description of Diameter, we devote this section to a comparison of RADIUS and Diameter as well as the co-existence of both during migration phases.
In the following subsection we list a number of improvements that Diameter as a AAA protocol offers over RADIUS.
Fail-over is defined as the process of forwarding all the pending requests with an agent to another agent, once a transport failure with the first agent is detected. For this to be possible, it is, however, required that the nodes have agreed on failure support by setting up a flag in their Diameter messaging.
RADIUS does not define a standard fail-over mechanism, and as a result, fail-over behavior can differ between RADIUS implementations. Diameter, on the other hand, is more resilient towards transport failures and provides a well-defined fail-over behavior. Diameter supports application layer acknowledgements and specific watchdog mechanisms to detect lack of activity. Diameter fail-over mechanisms are defined in [AAATR3539]. A pending message queue for every peer is maintained at a Diameter node. Upon receiving a response, the corresponding request is removed from the queue.
Support of server-initiated messages is only optional in RADIUS [RAD3576] and this makes it difficult to implement features such as unsolicited disconnects ...