In Forefront TMG, when you publish a farm of Web servers that perform the same role or host the same content, you can enable high availability for inbound access by configuring Forefront TMG to control the load balancing among the servers in the farm. Load balancing ensures that requests are distributed evenly among the available Web servers, detects offline servers, implements failover, and maintains farm servers, without disrupting current endpoint connections.
Forefront TMG load balancing ensures that traffic is distributed evenly between farm servers, by using the following mechanisms:
- Round-robin mechanism—By spreading requests
from different IP addresses evenly among the Web farm members, the
round-robin mechanism ensures that user requests to a Web
application serviced by a Web farm are distributed evenly among
farm members that are online. This even spread is preserved during
failover. When failover occurs, servers that are not responding are
detected, and the load is distributed among the available
servers.
- Affinity—Forefront TMG ensures that, after a
user has been routed once to a particular application server, the
user continues to be routed to that server. Forefront TMG supports
two load balancing affinity types:
- Cookie-based load balancing, or session
affinity—User session is associated with the server. It is
recommended that you use session affinity whenever possible,
because it provides a more reliable affinity when a Web server is
restarted.
- Source-IP-based load balancing, or IP
affinity—Client IP address is associated with the server. This type
of affinity should be used in an Exchange RPC-over-HTTP publishing
scenario, where session affinity cannot be used because cookies are
not supported by the Outlook client application.
- Cookie-based load balancing, or session
affinity—User session is associated with the server. It is
recommended that you use session affinity whenever possible,
because it provides a more reliable affinity when a Web server is
restarted.