Affect of compression on speed?

[pshute]

I'm experimenting with increasing compression levels on our backups. If the backup log says the bottleneck is the source, does that mean increasing the compression level should have no effect on the backup speed?

I ran a test active full job of a 30GB server at Optimal, High and Extreme levels. Each job's log said the bottleneck was the source. Sizes and times:
- Optimal - 30GB vbk in 30:00 minutes
- High - 26GB vbk in 32:40
- Extreme - 24GB vbk in 34:19

Can I expect those job times to scale up if I include more servers in the job? Eg will our normally 13 hour main job with Optimal compression now take 14 hours with High compression or 15 hours with Extreme compression?

If so, why? If the server is still keeping up with the data coming from the source, why is the total time longer?

The tests were run during office hours, so server load might be what caused these time differences, but they are consistent with more compression taking longer.

[PTide]

Hi,

Would you provide bottleneck stats percentage, source proxy location and type, backup mode (direct SAN, proxy appliance, network block device), please?

Thank you

[pshute]

Bottleneck stats:
- Optimal - 4/07/2018 11:27:43 AM :: Load: Source 99% > Proxy 27% > Network 2% > Target 0%
- High - 4/07/2018 10:48:37 AM :: Load: Source 99% > Proxy 50% > Network 2% > Target 0%
- Extreme - 4/07/2018 12:05:02 PM :: Load: Source 99% > Proxy 52% > Network 1% > Target 0%

How do I identify the source proxy location information and backup mode? From what I remember, the backup server is a standalone machine backing up VMs on 3 hosts on a Nutanix box physically above it in the rack. The backup server is connected to the Nutanix box via multiple 1Gbps NICs.

Does this entry from the backup log help?
4/07/2018 11:34:42 AM :: Using backup proxy VMware Backup Proxy for disk Hard disk 2 [nbd]
Does [nbd] stand for network block device?

[PTide]

How do I identify the source proxy location information and backup mode?

You can check that in the job settings

Does this entry from the backup log help?
Does [nbd] stand for network block device?

Yes, that helps, and yes, it stands for "network block device". Could you check in "Backup infrastructure" - "Backup Proxies" if there are other proxies in you infrastructure?

Thanks

[pshute]

Backup proxy is set to automatic. If I click on "Choose", the only option is called "VMWare Backup Proxy". That's also the only one listed in Backup Infractructure/Backup Proxies. In case it's useful, the transport mode of that proxy is set to Automatic, with failover to network mode enabled.

The job mode is set to Incremental, if that's what you meant by mode. For the three tests, I manually ran Active Full.

[DeadEyedJacks]

This implies you are using the Veeam Backup management server as your sole proxy, generally this is not optimal. Setting up dedicated backup proxies will take this workload away from the management server and likely improve your environment's throughput and performance.

[Gostev]

The usage of NBD transport looks to be the bottleneck here... adding a hot add capable backup proxy may accelerate overall processing performance. If this is not an option, then doubling the number of concurrent tasks on the default proxy should help to reduce the backup window (but not the speed of individual VM processing).