Proxy server scoping sizing

[stevecous]

Is there and tools or calculation that can be used to scope the number of proxy server required for a given number of VM's? I realise this is a bit pye in the sky as the size of the VM's can vary and the repository will have a bearing on this.

Thanks

[Jfmoots]

Nothing "official" but describing your environment here will get some great replies from guys that have implemented and run the software. As you indicated, there's a lot of variables at play that can influence how many proxies you deploy. I've even seen cases of too many. Tell us about what you're dealing with.

ESX hosts, # of VMs, Total size of the VMs if you know it, avg. percent of daily changes in virtual disks, Production storage, backup storage. If you're crossing a WAN, let us know the capabilities of that connection.

[Gostev]

Best is to do POC install on a small part of infrastructure... gives the good idea on the above.

Also, don't be afraid to over provision in terms of number of proxies. You will thank yourself later, when some of them go down for whatever reason. Actually, you will not even notice that some of your proxies are done, because all your jobs will still complete within the backup window but if don't have extra proxies and same thing happens, you can guess the consequences.

[rbrambley]

Without knowing any specifics on the virtual infrastructure, think of scoping proxies in these terms:

- Are you building backup jobs or replication jobs, or both?
- are you installing more than 1 Veeam Console / GUI?

Each install must use it's own proxies - the service can not be shared

you HAVE to have at least 1 proxy at both locations for replication
you probably should have a separate Veeam server with it's own proxies for replication and a different Veeam server for backups with it's own proxies
add an extra proxy for redundancy or fail over if one is down

- what is your window for jobs?

Is one data mover (proxy) enough to cover your expected nightly rate of changed blocks across all your VMs?
are you going to use one big proxy for multiple concurrent jobs or several smaller proxies?
Or can you get away with running your jobs one after another with a single proxy?

- How will your proxies connect to the ESX / Hyper-V data stores where the VMs live? Or, will you use VMs as proxies or physical servers, or both?

HBAs for FC storage, iSCSI initiators, etc for direct san access will probably be a physical server proxy - these can be bigger and handle more concurrent jobs since their resources (CPUs and RAM) are not being shared from/with the hypervisors
Virtual Appliance mode when proxies are VMs
Or network mode as a last resort whether virtual or physical proxies

-What's your backup target? How will the proxies connect to it, how many can send data to it at one time?
-what's your physical and virtual networking look like?

So, believe it or not, that's a simplified, off the top of my head, way to think about proxy planning. As you can see, it's different for every customer and what's important may not always be "how many". It's tough to make "a tool" for all of these scenarios, and there is no best way to do it every time from Veeam's perspective. We have features to handle any scenario you could architect.

Reality is that you can start with a best guess, run jobs for a while, and easily add or remove proxies to tailor and tweak jobs later.

[eptc]

[merged]

Hi,
Is anyone aware of any tools which can help size a veeam B&R environment
Thanks
eptc

[foggy]

No such tools, but you can search this forum for similar discussions and get some ideas of what others do in the environments of different sizes.

[Desbalais]

[merged]

Hello,

I currently use Veeam Backup and Replication in several small infrastructures, always through physical Veeam servers.
Now I need to deploy Veeam on a large infra. Large for me, perhaps not for you...

Here are the main informations :
- vSphere 5.0 Enterprise Plus
- 1 Blade Center
- 8 Blades with 256 Go memory and 2 sockets (10 cores each)
- Hyperthreading not enabled, but we are thinking to enable it
- 250 virtual machines
- 10 TB of vmdk on SAN datastores

We would like to implement Veeam in virtual machines instead of physical servers, because we're in a stream to virtualize all our infrastructure and remove all physical servers. But I never used virtual Veeam server in a production mode, only for tests.
Of course, if such an infrastructure cannot be backed up through virtual Veeam servers, let us know.

Have you already implemented such a solution, or do you have some ideas about the sizing ?
- How many Veeam servers (proxies and repositories, anyway we'll use 1 console)
- How should be sized Veeam servers (how mani vCPU, how many memory)

Of course we could start small and increase infrastructure each time there is a new bottleneck, but we need to plan some purchases : Windows licences for Veeam servers, perhaps a 9th blade if requested Veeam infrastructure is large.

If you have some experience or some tool to size the infrastructure, I would apreciate.

Thanks for any help
Yvan.