Bad Performance coping Veeam files on REFS repository

[sandrola]

Following Infrastructure:

Veeam Backup & Replication 9.5 U3 Server VM on vSphere
Veeam Backup & Replication 9.5 U3 Proxy Server Physical with 2 Core / 36 CPU, 128 GB RAM, Windows 2016
Scale-out Repository with multiple 16 TB extent, REFS formatted / 64 K Blocksize , mounted over FC from HP 3PAR Storage

Backup form vSphere to the Repository (Reverse Incremental) are very fast (over 1 GB/sec) for active Full Backups. During the Backup we don't have problems on the Proxy (RAM / CPU).

BUT:
If I try to copy .vbk File (let say 100 GB) from one repository (V:\) to the other repository (W:\) on the same Proxy server I reach 50-60 MB/sec Speed. This indipendent of copy with Veeam File copy or on the Windows OS with copy/paste.
If I create a big file with fsutil and I copy it, I reach 1.3 GB/Sec. speed.

Is there a explication for this?

[Gostev]

VBK fragmentation due to fast cloning?

[sandrola]

Hello Gostev
Veeam use fastclone to recreate the reverse incremental Files. Ist there a tool o see how the file is fragmented?
My concern is that Tape backup are very slow (55 MB/sec with LTO 7). The question ist if this poor speed is related to the poor performance with the disk to disk copy.

[Gostev]

No, there's no such tool unfortunately. But I realized that perhaps you may also be affected by ReFS performance regression from May updates, so try waiting until July patch Tuesday - and see if the issue is still there after installing the update. Thanks!