Global snapshot clones are instantly writable directories cloned from snapshots. Clones can be created from any snapshot residing on the cluster itself or on a cluster that has a replication peer relationship with the local cluster. The cloned directory can be from any snapshot, such as a snapshot replicated from a third cluster to a remote peer. It cannot be from a subdirectory of an snapshot.
A global snapshot clone is created within a short time of usually around 10 seconds or less, after which it is immediately writable on the local cluster on which it was created. After creation of a snapshot, the data can be copied over from the cloned snapshot to the snapshot clone, or not. Either way, users accessing the snapshot clone can request to read, write and list from the clone. If data requested for reading and listing has not been copied to the clone, it is read from the cloned source on the remote peer.
Background Copying Performance and Bandwidth Limits
When you enable background synchronization for a global snapshot clone, the cluster copies data from the source snapshot to the clone in the background.
The availability of the clone and the performance of this background copy process are governed by the following behaviors:
Data availability during copying
Files in the clone are available for reading immediately upon clone creation.
Files become available for writing or replication only after the entire clone has been copied.
Bandwidth scaling and resource protection:
Copying data at high bandwidths consumes significant system resources, specifically CPU and I/O. To manage this resource usage, the background copy speed is restricted to a default limit of 100 MB/s per write module (wmodule). This will result in a throttling of background copy speed of ~2 GB/sec per CNode (for Gen5 or Turin based CNodes/EBoxes).
Because the number of wmodules in a CNode depends on the CNode hardware, the total copy bandwidth automatically scales based on the size of the cluster. This allows larger clusters to leverage their hardware for faster copy speeds while protecting smaller clusters from resource exhaustion.