| 摘要 |
A network-level controller 103 allows servers 105 and application software to operate in a fully redundant hot standby mode. An application being protected runs on a primary server 105a and a secondary server 105b. Controller 103 has total control over the transfer of data packets. Data requests from clients 101 are duplicated (mirrored) by controller 103 and routed to both servers 105a, 105b. Both servers respond to requests but only data received from main server 105a is routed back to the client with data from standby server 105b being ignored. Controller 103 monitors servers 105 and if a server or application fails then a failover to the standby server is performed. The client-side IP address of controller 103 is a virtual IP address and servers 105 are assigned separate IP addresses which controller 103 communicates with. When a failed server recovers, controller 103 manages all background data synchronisation. Mirroring, synchronisation and control are performed at the packet level, thus allowing high availability of software and servers not designed for hot standby operation.
|
