Distributed Storage Fabric and Data Protection
Layout and redundancy
A vDisk is the unit the hypervisor sees. DSF divides it into vBlocks and stores the pieces as extents spread across nodes, so no single node holds a complete vDisk. Redundancy is set at the storage container level: RF2 keeps two copies and tolerates one component failure; RF3 keeps three and tolerates two. Erasure coding (EC-X) trades some rebuild performance for capacity on colder data and needs a cluster large enough to stripe across nodes.
Snapshots and protection domains
Nutanix snapshots are metadata operations - the first snapshot of a vDisk is nearly free, and later writes go to new extents (redirect-on-write). A protection domain groups VMs and vDisks and applies a snapshot schedule. Consistency groups keep multi-VM snapshots consistent across the group, which matters for applications split over several VMs.
Replication and DR
Async DR ships snapshots to a remote cluster on a schedule, giving an RPO of minutes to hours. NearSync narrows that to seconds or minutes. Metro Availability stretches a cluster across two sites with synchronous writes and automatic VM restart at RPO zero, but demands low latency between sites and a witness. Leap orchestrates recovery plans on top of these mechanisms.
Traps
Candidates confuse replication factor with RAID (there is no RAID group to rebuild), assume a snapshot copies data immediately, or think erasure coding improves performance. Remember too that replication is configured per protection domain, not automatically for the whole cluster.
Rule of thumb
Count how many simultaneous failures you must survive, add one, and that is your replication factor.