Abstract: This paper presented a scheme of highly-available,controllable redundant Ethernet,which sends same data packets through two network cards to avoid losing information when packets lost in the network．A frame redundancy label is added before Ethernet user data,used for marking redundant and irredundant data and distinguishing different frame from redundant data,which can reach zero recovery time when network problems happen．Network diagnosis is introduced in the scheme,which monitors network’s state through the software in the nodes sending,receiving and dealing with UDP messages．Winsock SPI and API are used respectively for achieving redundant communication and network diagnosis,in order to keep transparent to the application layer as well as guarantee the network’s high availability and usability．Experiment shows that the LAN can tolerate multi-problems from nodes to switches,increase the data sent in the network and consequently achieve higher fault tolerance.

Key words: Double redundancy,Ethernet,High availability,Network diagnosis