Configuring Cache Array Routing Protocol
By default, Cache Array Routing Protocol (CARP) is enabled for all the servers in the array, for outgoing Web requests. That is, the CARP algorithm will store objects in any one of the member servers' cache. By default, CARP is disabled for all incoming Web requests.
When a member server of an array determines that the requested object is not in its cache, it sends the request to another member server, using the destination server's intra-array IP address. Typically, this is the same Internet Protocol (IP) address that downstream clients and Microsoft Internet Security and Acceleration (ISA) Server computers use to communicate with this ISA Server computer. Because this value must be replicated to all ISA Server computers in the array, it is recommended that you do not change this value. For instructions on configuring a server's intra-array IP address, see Configure intra-array communication.
You can configure the member servers so that different servers have different loads. For example, if one server in the array has a disk four times as large as all the other member servers, you can configure that server to receive a proportionate amount of the cache load, by configuring its load factor. The load factor determines how to divide the load among members of an array. Changing this value increases or decreases the load on an ISA Server computer.
For more background information on CARP, see Cache Array Routing Protocol. For instructions on enabling CARP, see Enable Cache Array Routing Protocol.
For instructions on configuring the load factor, see Configure the load factor.
