Unfortunately, Microsoft hasn't brought any new logic to how its cluster technologies are named. To avoid mass confusion, you must realize that Microsoft offers two distinct clustering technologies. Each is designed for a different purpose, each is more or less incompatible with the other, and the two have nothing in common?yet they're both called clustering. They are
Cluster Service? This clustering is designed for high availability. Essentially, between two and eight servers function entirely on their own but are also capable of taking on the tasks of another server in the cluster should one fail. This behavior is similar to how most offices work: Joe and Bob normally have their own job tasks. However, when Joe goes on vacation, Bob gets to wear both hats and perform Joe's job tasks along with his own. In Windows terms, Joe and Bob are nodes in a Cluster Service cluster.
Network Load Balancing? This clustering is designed for high availability and scalability. Each server in the cluster?up to 32 are possible?is completely independent but performs the exact same work as the other servers. Work is more or less evenly distributed between them. In office terminology, this is similar to a customer service call center. Each call center employee can handle the same tasks, and the phone system directs incoming calls to the next available representative.
You'll find the Cluster Service most often used to cluster enterprise back-end servers, such as Exchange Server computers or SQL Server computers. Generally, companies build so-called active-active clusters, in which each computer is an independent, fully functional server. Should one server fail, however, another server in the cluster picks up the failed server's work. You can also create active-passive clusters, in which one server performs useful work and the other simply waits for the active server to fail. At that point, the passive server seizes control and becomes active, performing useful work.
NLB, on the other hand, is most often used in Web server applications. Large groups, or farms, of Web servers are configured to use NLB. Each server in the farm (or cluster) can perform the same work, and incoming user requests are distributed to the server that's least busy at the time. If the farm becomes overworked, new servers can be added at any time. NLB provides high availability because, if a server in the farm fails, other servers are already performing the same work. Therefore, users are able to continue working with the farm, if not with a particular server.