2.2. Components of a Cluster

2.2.1. Cluster Nodes

Cluster nodes do the actual work of the cluster. Generally, they must be configured to take part in the cluster. They must also run the application software that is to be clustered. Depending upon the type of cluster, this application software may either be specially created to run on a cluster, or it may be standard software designed for a stand-alone system. Turbolinux Cluster LoadBalancer 10 allows the use of software written for stand-alone systems. Configuring the software to be used within the cluster is usually pretty straight-forward.

We will sometimes refer to cluster nodes simply as nodes, servers, or server nodes.

2.2.2. Cluster Manager

The cluster manager divides the work amongst all the nodes. In most clusters, there is only one cluster manager. Some clusters are completely symmetric and do not have any cluster manager, but these are more rare today. They require complex arbitration algorithms and are more difficult to set up.

In Turbolinux Cluster LoadBalancer 10, the cluster manager is referred to as the Advanced Traffic Manager, or ATM. Turbolinux Cluster LoadBalancer 10 provides fail-over for the ATM so that there is no single point of failure. If the primary ATM goes down, a backup ATM will be able to fill in and take its place.

Note that a cluster manager may also work as a cluster node. Just because a system is dividing the work does not mean that it cannot do any of the work itself. However, larger clusters tend to dedicate one or more machines to the role of cluster manager, because the task of dividing the work may take more computational power. It also makes it a bit easier to manage the cluster if the two roles are isolated.