J.-M. Kim, S.-J. Bae, S.-H. Jeon, and H.-Y. Kim (Korea)
cluster computing, Linux Virtual Server(LVS), heartbeat, High Availability (HA), cascading failover
In this paper we propose a cascading high availability system for internet server clusters. Our high availability cluster system can be configured with a cascading failover mechanism with a primary/backup node to eliminate the load balancing system itself as a single point-of failure. There are two major factors that are building the cluster system for cascading with primary/backup node, high availability and faster response time. The failover within a cluster is handled through pre-defined policies that include methods such as electing backup node. The service nodes except backup node monitor primary and backup nodes. If primary node is down, pre-elected backup node takes over primary node, and then real nodes elect new backup node at the same time. Therefore we support efficient high availability to prevent delay of services from electing backup node and this protects clusters from multiple node failures.
Important Links:
Go Back
