Content for  TS 28.104  Word version:  18.3.0

This MDA capability is for failure prediction.

There are multiple sources of faults which may cause the 5G system to fail to provide the expected service. These faults and the associated failures need extensive troubleshooting. In order to reduce network and service failure time and performance degradation, it is necessary to supervise the status of various network functions and resources, and predict the running trend of network and potential failures to intervene in advance. These predictions can be used by the management system to autonomously maintain the health of the network, e.g. speedy recovery actions on a network function related to the predicted potential failure. Due to the fact that failure prediction could depend on the existing alarm incidents and relevant historical and real-time data (performance measurement information, configuration data, network topology information, etc.), there is a possibility for MDA to be used in conjunction with AI/ML technologies and model training to predict potential failures. In order to avoid the occurrence of failures and abnormal network status, it is necessary for consumers of analytics to obtain the required details of potential failure and the corresponding degradation trend (abnormal KPI, performance measurement information, possible alarm type, fault root cause, etc.). Therefore, MDA, may in conjunction with AI/ML technology, be required to obtain basic health maintenance knowledge (e.g. the relationship between the failures or potential failures and the related maintenance actions) through predefined expertise or model training, so as to effectively predict potential failures. The basic health maintenance knowledge could be updated with feedback. Besides the MDA capability to obtain basic health maintenance relationships between the service failure and related potential failures at network levels, the MDA capability for failure prediction may take role of coordination in cross domain. When MDA capability takes role of coordination in cross domain, the MDA capability for failure prediction can collect analytics output of fault prediction from single domain management and provide recommendation actions accordingly. If necessary, MDA could also provide corresponding recommended actions for failure prevention.

This MDA capability is for service failure recovery.

There are multiple sources of faults which may cause the 5G system to fail to provide the expected service. The potential management actions to support recovery of network service are some operational activities, such as switch to a redundancy NFs, modify configuration(s) of NFs based on different scenarios. When a service interruption disaster occurs (e.g. massive call disconnections), it is important for MDA to provide analytics to suggest management actions to quickly recover the service while avoiding causing other problems (e.g. signalling overload) during the recovery. The analysis of failure recovery can be used by the management system to recover from service interruptions in an orderly manner. As an example, in case that the service interruption occurs, the MDA can provide analysis of the possible recovery plan and the recommended actions for service recovery (e.g., update NF, change configuration of core network NF etc.).