* Connection to Cassandra takes a big amount of time, therefore it makes
sense to try the connection as soon as possible (even if you will not use
it due to just using volatile objects). In case the connection is not
possible (because the cassandra cluster is not started yet), then it just
silently continues... until a make_persistent, get_by_alias, ... is issued,
and then it will retry the connection (but loosing any performance gain...)
[this case, where you want you use persistent objects but Cassandra cluster
is not yet started... is really strange and should not be the usual case]