| 摘要 |
Methods and associated apparatus for rapidly activating a standby server (204) to take over for a failed server (202) in a redundant server network environment. The methods of the present invention configure the NIC of the redundant (standby) server (204) to receive information having a MAC address corresponding to the main server in addition to its own unique MAC address. Multi-cast features common to commercially available NICs may be used for such configuration. The standby server (204) may normally process information addressed to its unique NIC MAC address and the associated logical address (e.g., IP address) while also monitoring (receiving) information addressed to the main server's NIC MAC address. When the standby server (204) takes over for the failed main server (202), it begins responding to packets received for the failed server's NIC MAC address. In the preferred embodiment of the invention, the standby server (204) reconfigures its NIC MAC addresses to receive packets addressed to the main server (202) only upon detection of a failure in the main server (202). Responses are generated by assuming the logical address identity of the failed server (202) but with the lower level (MAC) address uniquely assigned to the redundant server's NIC. Such replies cause clients (106) to immediately update any address cache (e.g., ARP tables) (102) without needing to await delays associated with error recovery timeouts. In addition, the present invention is applicable to a wider variety of network topologies, including ring topologies such as Token Ring and FDDI, as compared to prior techniques because the physical address generated by responses from the redundant server are consistent with its unique MAC address. |
