We have one set of similar backup jobs where we are duplicating full backups from disk to disk via SMB for archival purposes. Duplicate jobs of full backups of our hyper-v instances demonstrate reasonable performance of 450MB/min across our VPN. The problem is "any" direct backup jobs that backup non-hyper-v servers have absolutely horrifyingly low "duplicate" performance of 40-50MB/min to and from the same storage. This issue has persisted for several years. I opened a support ticket or two during this time but always end up closing the tickets after a couple weeks as the support tech's knowledge beyond the UI of Backup Exec is frustratingly limited.
To isolate the issue, I have verified it does not matter if we're duplicating jobs in dedup storage or local disk storage, either have the same exact issue. I do not believe the issue is related to the source location of the full backups. I believe the issue is related to the remote nature of the SMB share and Backup Exec's underlying processes when duplicating direct non-hyper-v full backups.
Our environment includes the Backup Exec server which is Server 2019 to a remote Server 2019 file share. All drivers and updates installed. The underlying disk storage has performance in excess of 1TB/sec and the network interface is 40G Mellanox on a 100G Cisco backbone. We have 100mb/sec of guaranteed internet bandwidth dedicated for this purpose and to avoid excessive packet fragmentation across the VPN and internet, limit packet sizes to 1418 bytes which was determined by pinging the remote resource with a 1419 packet size where fragmentation begins. SMB performance of this link hovers around the 450MB/min level which matches the Hyper-V duplicate jobs and is what we expect.
The problem is any "duplicate" job that duplicates full backups of servers that are backed up directly is always limited to 40-50mb/sec. The data we are duplicating is universally the same. i.e. Backup Exec backs up a non-hyper-v server like a Linux workstation, Windows workstation or non-hyper-v Windows server and we are dupicating these.
Alternatively, as a workaround, we've had to abandon Backup exec for the duplication of these jobs use robocopy to mirror the backup exec data files directly. At least with this solution we have compression and resume options and can utilize all 100mb/sec internet bandwidth we have allocated for this. This is not desirable as we have to turn off the lockdown server and constantly monitor robocopy and of course, this is all manual with no reasonable way to handle reporting or job status within Backup Exec.
Beyond this, for future versions of Backup Exec, it would be immensely helpful if "duplicate" jobs could resume where they left off instead of being required to start over. It would also be helpful if Backup Exec could utilize compression for the transmission of all network data and be able to utilize multiple network adapters like SMB3 does.
The question I have is what is Backup Exec doing that is limiting these specific "duplicate" jobs to 40-50MB/min and how can we resolve this without resorting to robocopy?