For the purposes of the present document, the terms and definitions given in TR 21.905 and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in TR 21.905. Service Interruption:

For the purposes of the present document, the abbreviations given in TR 21.905 and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in TR 21.905. GRUU LIA LIR MAA MAR OA&∓M O&M OPEX SAA SAR UAA UAR

This clause contains a list of the requirements for the solution provided in this study. The general goal is to have a set of procedures ensuring that the impact of the service interruption of a node is limited to the loss of the capacity of that node for the time that it is out of service, plus some additional signalling in order to perform the take over by other network elements with the same function.

There are network elements that hold permanent data. There should be methods to ensure that the information in these network elements is not lost even in disaster events, such as a complete site crash. For this reason, this study assumes that the redundancy provided for these network elements makes unnecessary additional network procedures for restoration of permanent data. On the other hand, temporary data will be considered in these procedures. For the nodes that don't handle permanent data, the assumption is that their memory is affected if an outage occurs and the information related to some of the users may be lost.

Interruption of established sessions is considered an acceptable consequence of the failure service interruption of one of the network elements in the session path. This means that the restoration of session data does not need to be analyzed in this study. Means may be taken by implementations outside of this study to enable established sessions to be restored or maintained. The accounting of IMS sessions is already based on principles (interim accounting as described in TS 32.260) that ensure that the charging of this interrupted sessions is also terminated.

The restoration procedures could involve some steps that take place during the establishment of sessions after the outage has occurred (i.e. after the network element that failed returns to normal functioning or another network element takes over). If that is the case, the increase of the session establishment time should not be significant (it should still have a high probability to remain within the acceptable levels for the end user).

In order to reduce OPEX and the time required to restore the network, the need for manual intervention should be minimized. This implies that the procedures should be triggered by network signalling events and O&M steps should also be minimized.

Loss of service refers to the state in which session origination attempts by the user or session termination requests to that user fail while the UE appears to be registered and also when the network does not respond to registration attempts. Ideally the proposed solution should avoid this kind of loss of service altogether and ensure that requests are terminated correctly in all cases. If this is not feasible, then the time of loss of service for the user should be minimized.

In the solutions provided it needs to be taken into consideration that network element failures tend to occur in situations when the signalling is overloaded. If that is the case, restoration procedures that involve a high level of messages (e.g. triggering re-registrations for all the UEs controlled by a P-CSCF or S-CSCF) should be avoided. Such kind of procedures could result in further problems in other network elements and provoke a domino effect of subsequent failures.

The solution provided should be such that it allows the recovery of the network to a situation where the load is balanced between network elements performing the same function.