Your network consists of a single Active Directory domain. All servers run Windows Server 2008 R2. You plan to publish a Web site on two Web servers.

You need to deploy an availability solution for your Web servers that meets the following requirements:

– Supports the addition of more Web servers without interrupting client connections
– Ensures that the Web site is accessible even if a single server fails

What should you do?

A.
Configure a failover cluster.

B.
Configure a Web garden on each Web server.

C.
Create a Network Load Balancing cluster.

D.
Create two application pools on each Web server.

Explanation:
MCITP Self-Paced Training Kit Exam 70-646 Windows Server Administration:

Windows Web Server 2008

Windows Web Server 2008 is designed to function specifically as a Web applications server. Other roles, such as Windows Deployment Server and Active Directory Domain Services, are not supported on Windows Web Server 2008. You deploy this server role either on a screened subnet to support a Web site viewable to external hosts or as an intranet server. As appropriate given its stripped-down role, WindowsWeb Server 2008 does not support the high-powered hardware configurations that other editions of Windows Server 2008 do. Windows Web Server 2008 has the following properties:

The 32-bit version (x86) supports a maximum of 4 GB of RAM and 4 processors in SMP configuration.
The 64-bit version (x64) supports a maximum of 32 GB of RAM and 4 processors in SMP configuration.
Supports Network Load Balancing clusters.

You should plan to deploy Windows Web Server 2008 in the Server Core configuration, which minimizes its attack surface, something that is very important on a server that interacts with hosts external to your network environment. You should only plan to deploy the full version of Windows Web Server 2008 if your organizations Web applications rely on features such as ASP.NET, because the .NET Framework is not included in a Server Core installation.

Configuring Windows Network Load Balancing
While DNS Round Robin is a simple way of distributing requests, Windows Server 2008 NLB is a much more robust form of providing high availability to applications. Using NLB, an administrator can configure multiple servers to operate as a single cluster and control the usage of the cluster in near real-time.

NLB operates differently than DNS Round Robin in that NLB uses a virtual network adapter on each host. This virtual network adapter gets a single IP and media access control (MAC) address, which is shared among the hosts participating in the load-balancing cluster. Clients requesting services from an NLB cluster have their requests sent to the IP address of the virtual adapter, at which point it can be handled by any of the servers in the cluster.

NLB automatically reconfigures as nodes are added and removed from the cluster. An administrator can add and remove nodes through the NLB Manager interface or the command line. For example, an administrator might remove each node in turn to perform maintenance on the nodes individually and cause no disruption in service to the end user.
Servers within NLB clusters are in constant communication with each other, determining which servers are available with a process known as heartbeats and convergence. The heartbeat consists of a server participating in an NLB cluster that sends out a message each second to its NLB-participating counterparts. When five (by default) consecutive heartbeats are missed, convergence begins. Convergence is the process by which the remaining hosts determine the state of the cluster.

During convergence, the remaining hosts listen for heartbeats from the other servers to determine the host with the highest priority, which is then selected as the default host for the NLB cluster. Generally, two scenarios can trigger convergence. The first is the missed heartbeat scenario mentioned earlier; the second is removal or addition of a server to the cluster by an administrator. The heartbeat is reduced by one half during convergence. A less common reason for convergence is a change in the host configuration, such as a host priority.