| 摘要 |
A method for increasing the availability of a first server included in a computer cluster when a second server fails. Each server in the computer cluster has an associated mass storage device and can process requests from any network device in the computer cluster. Data is mirrored between the mass storage devices of the servers so that each server's mass storage device has a complete copy of all computer cluster data. Data mirroring takes place across a dedicated link, which reduces congestion on the rest of the computer cluster. When the first server detects a loss of communication from the second server, the first server determines if the loss of communication is a result of a malfunction of the dedicated link. If the dedicated link has failed, the first server discontinues operation to avoid writing data to its associated mass storage device, which cannot be mirrored due to the loss of communication. If the dedicated link is operational, the first server continues operation. In either case, since each server can process requests from any network device and each server has a complete copy of all the network data, the computer cluster continues to be available for use even after a server is shut down.
|
