Backup Proxy Recommendations

[jlatulippe]

Good day,

I am very new to Veeam B&R (v9.5) and have been doing pretty well in setting up our backups so far, but have a mental "block" when it comes to the backup proxies and need some advice (and perhaps clarity) from the more experienced among us!

Infrastructure:

We have two Hyper-V servers (2012R2) currently housing vm's that are being backed up; one with 4 CPU's and one with two (so taking advantage of Essentials (Enterprise Edition) - backing up 12 VM's including Exchange 2010 and SQL Server 2008R2.

Currently the backup proxy is the backup server (default), and working pretty well. Its a Dell R610 w/two CPU's and 64Gbytes of RAM. The Backup Repositories are two Dell 910 servers with 4 CPU's and 128GBytes of RAM per server. They each have 4TB volumes on which I am storing the backups (RAID10 array comprising 8x1TB drives). I have already ordered 16 additional 1TB drives (8/server) to create additional storage for longer retention/restore points.

The Veeam backup server, Hyper-V servers and repositories are connected on the same physical LAN and are across two Ethernet switches. I am actively working on increasing our Internet bandwidth and once completed I will be setting up replication of these vm's to our DR site via VPN connectivity across the Internet.

Now, my question(s):

Q1.) Am I best to offload the Backup Proxy service to these repositories, as they have better specs than the backup server and are only used as Veeam Repositories within our production network.
Q2.) If so, do I setup one or both as proxies?
Q3.) If I setup both as proxies does Veaam "know" which one to use for specific vm backup jobs?

Thanks so much for any guidance you can provide...

- Joe

[foggy]

Hi Joe, what are the bottleneck stats for your current jobs? Depending on this information you may decide whether you need to offload processing to off-host proxies or not.

[jlatulippe]

Thanks Foggy.

The bottleneck for all jobs states its the "Source". Following are the stats from all jobs (each job consisting of a single vm backup, staggered over our backup window):

Load: Source 99% > Proxy 2% > Network 12% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 2% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 1% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 98% > Proxy 10% > Network 27% > Target 3%
Primary Bottleneck: Source

Load: Source 99% > Proxy 5% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 5% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 3% > Network 34% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 2% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 2% > Network 2% > Target 0%
Primary Bottleneck: Source

Load: Source 99% > Proxy 2% > Network 3% > Target 0%
Primary Bottleneck: Source

Load: Source 84% > Proxy 7% > Network 71% > Target 1%
Primary Bottleneck: Source

Load: Source 99% > Proxy 3% > Network 12% > Target 0%
Primary Bottleneck: Source

Looks like the bottleneck is the source, our Hyper-v servers I assume this means. But what does this actually mean? The physical servers are the highest latency in getting the backups from source to target? How do you improve efficiency? Does it even need improvement as I am really not seeing any issues, just want to make sure I have everything optimally configured...

Thanks all,

- Joe

[foggy]

There will always be a bottleneck somewhere in the entire processing chain, so if you're satisfied with the overall performance and meet your backup window, there's nothing to worry about. However, if you'd still want to improve performance, in your case you indeed can offload the processing from the Hyper-V hosts by switching to off-host proxies. Please review the corresponding user guide section for details on how to configure them.

Btw, what is the reason of the "one job per VM" type of setup?

[jlatulippe]

Thanks for your reply. Yes, all seems to be going smoothly. The reason for the one vm per job was due to a gradual deployment I suppose. I created a single job which included one vm as a starting point, then when I was comfortable with its progress I created a second, and then a third. Having a sufficiently long backup window I figured I could better manage the start/end times with single vm jobs, as well as keep the load on the Hyper-V hosts down. I am now reading that grouping VM's in a single job is better for deduplication, so reconsidering my initial choices. Do you have any recommendations when it comes to grouping vm's into a backup job?

Thanks again,

- Joe

[foggy]

Due to parallel processing, Veeam B&R is able to optimally utilize the available resources to provide the shortest the backup window. Typically it is recommended to group VMs by guest OS type, here's an existing thread regarding that.