2.3. Types of Clusters
As you saw in the previous section, the definition of a cluster is pretty loose. So loose in fact, that there is some confusion about how differing technologies can all be referred to as clusters. The fact is that clusters can be implemented for several different reasons.
The most common reasons to create clusters are to pool CPU resources, balance a workload among several machines (load balancing), create high system availability, or provide a backup system in case the primary system fails (fail-over). These represent different types of clusters, although there is quite a bit of overlap.
Turbolinux Cluster LoadBalancer 10 can be used to implement high availability, load balancing, and fail-over. It does not provide shared processing in the usual sense of the term. Instead, it provides load balancing of network services. Each server receives incoming network service requests, processes the requests, and sends the reply back to the client.
 | Fail-over in this instance refers to moving a service to the stand-by node after the active server failed. Transaction fail-over is not supported by this product. |
2.3.1. Load Balancing
Load balancing is similar to shared processing, but there is no need for communication between the nodes. With load balancing, each node processes the requests it has been given by the cluster manager. The cluster manager will distribute the requests in some manner that attempts to distribute the workload evenly among all the systems.
2.3.2. Fail-over
Fail-over is similar to load balancing. However, instead of requests being distributed among all the cluster nodes, one system processes all the requests. Only when that system goes down will one of the other systems in the cluster take over.
2.3.3. High Availability
While it would be desirable to have all computers working all the time, the reality is that computers do sometimes go down. In some situations this is merely a nuisance, but in others it can be devastating. Therefore computer companies have devised methods of increasing the availability of systems. High availability is a method by which system resources are kept available as often as possible. Clustering provides a convenient way to do this. Instead of paying exorbitant costs for hardware redundancy, multiple systems can be clustered together to provide the needed resources. If one of the systems fails, the others can take over the workload.
High availability can be implemented with either hardware or software. Hardware systems are usually more expensive, but software solutions are generally not cheap either. The more reliability you require, the more you will end up paying.
Availability is often measured in percentage of uptime. A typical server may be up 99% of the time, whereas a system designed for high availability may be up 99.99% of the time. This is often referred to as "four nines" availability.
High availability can be achieved using either load balancing or fail-over.