Assuming that data needs to be replicated across servers on the hash ring.
After a key is mapped on the hash ring, the first servers that lie on the clockwise direction of the key store copies of the data
With virtual nodes, the first nodes on the ring may be owned by fewer than actual servers. So, we can choose to only choose unique servers while performing this replication
We can also take care to place replicas in distinct data centers to avoid single point of failure (by placing all replicas in a single data center)