TABLE OF CONTENTS Depending on the way alternative service paths are provisioned, service recovery can be broken into two classes: service protection and service restoration. Because most of the control plane aspects path computation and selection, advertising, signaling, and so forth are identical for both classes, the common term service recovery is used most of the time. Only when there is a need to highlight peculiarities of one of the two classes are the terms protection or restoration used instead.
In the early days of GMPLS, when the work on service recovery had just begun, there was a very clear distinction between service protection and service restoration. The former was defined as a class of service recovery where one or more alternative paths were fully established (that is, computed and signaled, with resources reserved, selected, and committed on all links, cross-connects programmed and prepared to switch traffic on all nodes) before a failure was detected. Service restoration was defined as a class of service recovery where no control plane actions related to the provisioning of alternative path(s) occurred until after failure had been reported to the deciding entity. In other words, in the case of service restoration, only when the deciding entity receives a fault notification is it supposed to trigger an alternative path setup starting from the path computation function.
The definition of service protection remains unchanged. However, it turned out that there could be many other ways to provision alternative paths...
