A lot of array vendors are supporting redirect-on-write snapshots and deduplication/compression. When doing replication to a second array the data is already in deduped/compressed format and doesn't take up as much network bandwidth.
Snapshots are near instantaneous, whereas network backups take time depending on the size of the master or size of the incremental.
Network backups require a full copy of the data, snapshots do not.
From this, I believe array snapshots are better in terms of space efficiency and RTO for backups and restores of files and servers. Is there something I'm missing here? With snapshots, does data have to be hydrated to do file-level searching and indexing, for example?