The monitoring of network resources and services, isolating and identifying problems when failures occur, and fixing them by sending technicians to the sites most of the time or downloading certain configuration files remotely to fix configuration related problems are common in the telecommunications industry. This is costly. In order to reduce the operational cost, it is necessary for a network isolating and identifying problems by itself and fixing them, and having technicians at the failure site only when there is a single point of hardware failure. This paper introduces the concept of a self-managed network to identify network problems during failures and repair them, in addition to self-configurations of network resources and services. Self-managed Network Element (sNE) architectures and Network Management System (sNMS) architectures for centrally managed networks are described. A hierarchy among repairing entities is defined. An in-band message format for Metro Ethernet networks is proposed for the fault management communication.