To hopefully provide some meaningful contribution to the conversation:
I’d always recommend using QoS at your egress firewall/router, over Veeam’s features.
Firstly as it’s aware of bandwidth utilisation across all link traffic. It means you can offer higher throughput during quieter times of link utilisation, but allow it to slow down & free up bandwidth to other applications, such as email or VoIP.
Secondly, and I’ll hold my hands up here and say I experienced this behaviour on VB365 but can’t confirm if VBR does the same or different, when I used bandwidth throttling on VB365, it would throttle bandwidth based on a calculation of usage over time.
So if you said no more than 20Mbps, if I had a 100Mbps link, I’d see over a 5 second average that it consumed no more than 20Mbps, which is great, although in reality what I’d see was 1 second of total bandwidth utilisation (the full 100Mbps link) and then 4 seconds of no traffic. This isn’t likely to be what you want as this would cause jittery traffic and poor experiences with real time communications. The details on this processing are here: https://helpcenter.veeam.com/docs/vbo365/guide/vbo_threads_and_limit.html?ver=60 but I can’t find anything equivalent within the VBR documentation to confirm what VBR does or doesn’t do.
It’s also detailed a bit further in this link: https://helpcenter.veeam.com/docs/backup/vsphere/setting_network_traffic_throttling.html?ver=110
You’ll find that throttling is only for backups, but not restores, which might not be your preference if you’re restoring some long term data from a year ago, but suddenly your entire WAN link is saturated for the duration of the restore (there are of course some scenarios where this is preferred too of course!). Finally network throttling is per VM disk, so you’ll still see some initial bursting of traffic at the beginning of the job as detailed in the link also.
I’m not saying that Veeam’s functionality is bad, it works for a lot of scenarios, I just like to apply flow control at the bottleneck, where we can see everything trying to navigate the bottleneck, and apply our logic there.
Thank you all for the informative responses.
