Configuring Network Load Balancing

  • Article

You define how the cluster load-balances client requests (and its other behaviors) by using the following Network Load Balancing parameters:

  • Cluster parameters (primary IP address, subnet mask, full Internet name, multicast support, remote control password, and confirm password), which specify the behavior of the cluster.

  • Host parameters (host priority, initial value, dedicated IP address, and subnet mask parameters), which define how each host functions within the cluster and in load-balancing.
    Host parameters are unrelated to the scenario for which you configure the Network Load Balancing cluster.

  • Port rules (port range, protocols, filtering mode, affinity, load percentage, equal load distribution, and handling priority parameters), which define how the hosts distribute the incoming requests on a port or range of ports.
    Port rules define distribution of client requests for each scenario and must match on every cluster host. If a server attempts to join the cluster with a port rule that is inconsistent with the rest of the cluster or is incorrectly specified, the server is not accepted into the cluster, and the current load distribution is unchanged. The cluster does not complete convergence while there is a host with a port-rule mismatch.
    For information about ports and a useful list of port assignments, see the appendix "TCP and UDP Port Assignments" in the Microsoft ® Windows ®  2000 Server Resource Kit TCP/IP Core Networking Guide .

For information about the basic concepts of Network Load Balancing parameters and about configuring cluster parameters, host parameters, and port rules, see Windows 2000 Network Load Balancing Help.

For information about scenario-driven guidelines to setting values, see "Implementing Network Load Balancing" earlier in this chapter.

Client requests that you do not want to load-balance are a special case. For information about how to prevent load-balancing of a class of client requests, see "Default Handling of Client Requests" later in this chapter.