When you add a new SQL instance to an existing cluster you specify the
virtual server name (it can be anything - "bert" for example).  Then you
say which of the cluster nodes you want "bert" to be able to run on -
there's pretty much no reason you'd include a node in a cluster and then
not allow a SQL instance to run on it (at least not in normal
circumstances).  You also define what you want the instance name to be
(eg. "ernie").  So the full name of you SQL instance would be bert\ernie
and it would be able to run on whichever cluster nodes you specified
during the SQL install (which, presumedly would be all of them).  If the
node it's running on fails, then it will start up on one of the other
possible owners (ie. nodes) in the cluster.  The virtual server name,
and in fact the entire SQL instance name (eg. bert\ernie), will remain
the same no matter what node the instance happens to be running on.  The
cluster group contains, among other resources, an IP address and a
network name (eg. bert), so those resources will "follow" the SQL
instance to whichever node it's running on - so the node it's running on
is transparent to the SQL users (they just always refer to it as
bert\ernie no matter where it is).

--
*mike hodgson*
http://sqlnerd.blogspot.com