Question on Veeam B&R + 1 Proxy and 1300 VM

[linux4guru]

Hi All,

I have a question surrounding this that our VMware Administrator brought up. When she was at VM World in SF, she meet someone from one of the event and this person said that he only have "ONE" proxy server and backing up 1300+ VMs and all backed up within the backup windows. My question to her was how is that POSSIBLE when per Veeam - the spec is 2 proc per jobs.

Is that even possible? Anyone out there running in this similar setup? I would like your input.

Thanks,

Thomas

[tsightler]

Technically, it's 2 CPU cores per job, so if you use a physical proxy with a lot of cores, you can backup a very large number of VMs with a single proxy. Also, you don't say what the "backup window" is, or the size of the VMs, this also has a huge impact on how many VMs. I had one client that had a single V6 server, running as a VM with 4 vCPUs that was backing up right at 1000 VMs in 12 hours (his backup window was 6pm-6am), but just barely. Splitting the load to 4 proxies got the backups done in ~4 hours, so there was a huge improvement, but he was technically meeting his window previously.

My normal rules of thumb are to plan for 50-100 VMs for each virtual proxy, and 250-500 VMs for each physical proxy, based on the CPU resources available, this gives you some headroom in case you have proxies that fail or get stuck running one job for very long, but honestly, physical proxies are growing in CPU horsepower so I'm seeing customers get by with less physical proxies. I always like to approach it from an N+1 perspective if possible though.

[linux4guru]

Thanks for replying... I like to use physical proxy but that didn't do too well for me. Was getting 4Mb/s on backup on all VMs. Switching to proxies fixed our backup performance issue. I didn't say the backup windows or the sizes of the VM because I didn't get that answer from our VM admin. Your info will help me a lot.

[chrisdearden]

4 MB a second is quite slow.was the physical proxy using nbd or direct San?

[dellock6]

hi Thomas, I attended that session too and the user case was about an environment based on Veeam V5, so without role separation coming with V6 that kind of design was somehow mandatory. They used physical veeam servers running side by side, each with dual socket westmere processors and local storage used as repository, and they do not use a single server. VMs were divided in several jobs and each server was responsible for a defined set of VMs. It works, but with V6 we can create a more distributed and balanced job design.
Choosing between physical or virtual proxies depends on the environment, some situation like FC storage are more suited for physical backup servers but is not mandatory even in this case, there are many elements to be considered as always. Having multiple proxies, as Tom stated, is a good design choice regardless going physical or virtual, both for performances and resiliency of the backup infrastructure.

Luca.

[linux4guru]

The physical proxy was using nbd and switching to vm proxy fixed the issue - now the vproxy is a vm that lives in the netapp and have a dedicated 10GbE to the Exagrid also on the 10GbE connection. Performance are very fast now.

4 MB a second is quite slow.was the physical proxy using nbd or direct San?