Depending on what features and level of failover you want you could choose to just use DNS to round robin between servers. This works by alternating servers through DNS resolutions, giving out the canonical names of a group of servers in a rotation. This method of load balancing doesn't account for server failure, so if a server stops working users accessing those services will witness the failure every time you move through the rotation.
The following solutions are more robust and would accomplish similar goals as NLBS:
- Linux Virtual Server
The Linux virtual server is a transparent load-balancing mechanism that can leverage a group of physical servers to create one virtual server.
- Red Hat Cluster Suite
The Red Hat cluster suite is an application and load balancing package that is sold by Red Hat that can support up to 8 nodes.
- You might also find that the High Availability Linux Project might be of some use to on understanding the principals associated with high availability and redundancy on Linux.
This was first published in September 2005