Hi,
we are making virtual machine backups using Veritas NetBackup 8.1 with their integrated VMware backup solution (through VADP) on a Nutanix (VMware vSphere 6.5).
It does’nt make any difference if we are using
- network block device (NBD; backupserver is sitting outside Nutanix and is transferring the data through 10GBit/s network; which is the recommended transport mode regarding to best practice guide from Nutanix and Veritas) or
- hot-add (backupserver is sitting on a VM inside Nutanix)
We are getting only a maximum of 30MB/s throughput (also against a nulldevice, so, target device or network is not the bottleneck) per backup job.
Of course, this is much too slow, if backing up a single big VM (some TByte).
When running multiple (for instance 10 VM) backups at the same time, the overall throughput increases, but not linearly. In other words: with 10 VMs at the same time we see about 160MB/s, instead of expected 10 times 30MB/s = 300MB/s.
Is there something known about that? Could it be, that there is a limit regarding to a single stream to prevent the storage from heavy load and is there a chance to change that behaviour?
It seems that NBU is not the reason for the issue, but my hope is, that someone already experienced and solved this.
Greetings, Matthias
