Have a customer using backup files as the source for their replication, and pointed out that the VM disks are processed sequentially, as stated in the user guide
Replica from backup processes disks sequentially, not in parallel
Just wanted to verify my understanding. So customer has 3 Jobs, with multiple VMs, and each VM has multiple disks
Job 1
VM 1 - Disk 1 / Disk 2 /Disk 3
Job 2
VM 2 - Disk 1 / Disk 2 /Disk 3
Jobs 3
VM 3 - Disk 1 / Disk 2 /Disk 3
the proxy (main repo server in this case) has 32 tasks (32 cores CPU) available, so if we run the replication jobs together , only the 1st disk in each job and VM is processed at a time, so will only consume 3 tasks?
are you you saying VM1, VM2 and VM3 will be processing in parallel, but only one disk at a time, so Job 1 would only consume a max of 3 tasks (1 per disk). And in Job 2 VM4 and VM5 (only 1 disk at a time) only a max of 2 tasks consumed? A total of 5 tasks?
If so , then if the 32 core server would have 27 task sitting idle, waiting?
But if we switched these jobs to use replication from snapshots, then we'd be processing all 9 disks at the same time ( 9 disks = 9 tasks)?
are you you saying VM1, VM2 and VM3 will be processing in parallel, but only one disk at a time, so Job 1 would only consume a max of 3 tasks (1 per disk). And in Job 2 VM4 and VM5 (only 1 disk at a time) only a max of 2 tasks consumed? A total of 5 tasks?
Correct, if you use Replica from backups.
But if we switched these jobs to use replication from snapshots, then we'd be processing all 9 disks at the same time ( 9 disks = 9 tasks)?
Yes.
I run a short test in my environment with a single job and 6 VMs. Just to verify both scenarios.
- Replica from Backup: 1 disk per VM processed
- Replica from Source VM with snapshots: multiple disks per VM processed