Based on a question asked in Lotusphere Gurupalooza, I thought it may be
useful to flesh out some details on load balancing the Domino HTTP service
and any applications that you want available over a browser but require
fault tolerance.
Clustering Domino servers in a notes environment is very straightforward.
 As two Domino servers are clustered, the are constantly sharing any
alterations to databases with each other.  This uses the cluster replicator
task (clrepl) to "push" changes in databases to other members
of the cluster.  In the case where one of the servers becomes unavailable,
or reaches its peak load, the Notes client is "cluster aware"
out of the box and will
(a): automatically detect that the server it is interacting with is a member
of a cluster and
(b): locate and move to the database on another server that is a member
of that cluster.

The cluster method in Domino far outweighs hardware based clustering technology
as it is not dependant on similar hardware, similar build versions or even
similar OS underlays.  This feature is a massive advantage to Lotus
as it "simply works" using the network level replication technology
that Lotus have been working with for years.  Of course, when the
client is not  notes , but a browser, you have to consider that these
clients are not "cluster aware" and will not redirect to another
server.  How do you address this?

Setting up a cluster in Domino is EASY. The hardware doesn’t have to be the same, the Operating system doesn’t have to be the same, the release of Domino doesn’t even have to be the same. It is a model of redundancy that obeys the OSI model fully. Many of my M$ buddies often argue the pros and cons of Domino -v- Exchange but throw in the towel when it comes to this feature. Because it is easy, sometimes administrators don’t bother to learn the monitoring details of clustering or how to tell if its performing ok.