112 CHAPTER 6 Hyper-V and high availability Clustering is done by using technology to group two or more servers together into a single functional unit. Traditional clustering techniques require the individual servers, or "nodes," to be physically identical--that is, they need to have the same hardware configuration, memory size, CPU speed, etc. They also need to be identical in software configuration, with the same OS and (ideally) patch levels, and the same applications installed. In addition, all nodes of a cluster must have access to the same data storage, at the same speed; this is typically done with a high- performance SAN. There are two common types of clusters: fail-over and load-balanced. Fail- over clusters consist of a single node that typically handles all of the client requests, called the Primary node, and one or more nodes that are largely inactive unless the Primary node goes offline; these are called Secondary nodes. In a load-balanced cluster, all of the nodes participate actively in serving client requests. In most cases, a load-balanced cluster can also serve as a fail-over cluster, since one or more nodes of the load-balanced cluster can typically fail without the other nodes being impacted. In any cluster, two major challenges present themselves: determining the status of a node member (particularly in fail-over clusters), and determining which node of a cluster currently controls a clustered application and its data. The first challenge is met with a heartbeat network, which is typically a physically separate set of network cards that communicate a signal, or heartbeat, to determine the status of each node. Data ownership is tracked by a data partition called the Quorum. The Quorum is a separate partition from the shared data, that also needs to be equally accessible to all nodes in a cluster. The quorum tracks which node is the owner of a given set of applications or data. A simplified diagram can be seen in Figure 6.1. MICROSOFT FAIL-OVER CLUSTER AND NETWORK LOAD BALANCING Microsoft has built into Windows Server since 2000 the ability to create a fail-over cluster in their Enterprise and Datacenter editions. The process is greatly simplified in Windows Server 2008, utilizing a wizard-based approach to validation and configuration of the nodes. The validation por- tion of the setup wizard checks your nodes as well as the available storage and network connections to ensure that the infrastructure will support the cluster, as well as checking the applications to ensure that they can support a clustered environment. It will also confirm that the server hardware is sufficient to support a clustered environment, before performing the steps necessary to set up and configure the cluster.