Triple or Multiple Architecture and Method for High Availability Guarantee on Edged AI Service

摘要

Distribution and synchronization of AI 　 service model between Center AI and Edge AI service, Center transmission of AI 　 service learned at each edge and re-learning using it, redistribution method according to AI 　 service improvement, and mutual change detection when AI service changes, if necessary In the event of redistribution, edge or center AI service interruption, an immediate recovery and synchronization method is provided. In order to monitor the status of the center AI and each edge AI, the AI service module and the independent coordinator module are linked with each AI service and then each AI service is The status is to share status information between coordinator modules. Accordingly, it is possible to detect the state of each edge or center in real time, and to restore the service in the most effective way to ensure high availability. In addition, by sharing the AI learning results of each edge without data sharing, the continuous AI 　 service improvement effect can be redistributed to all edges to continuously maintain the improved performance.