Couchbase Cluster Replica and Cluster node failure

Can anyone tell me what are the advantages of more than one bucket replica? How do you cater for more than one server failure in the cluster? Is it use more than one replica? If so how many servers in the cluster do you need and how many replica’s to support multiple server failures? Whilst this link provides info on a single replica it does not answer my questions above. Any assistance would be much appreciated.