Planning an evaluation / trial

[HighLevel]

Hello,

New to Veeam Backup and Replication and as we have virtualized most of our server estate we are looking at replacing our legacy (Non VM aware) backup product and want to try an evaluation of Veeam.
Just started looking at product guides but any advice around how to plan a potential deployment would be appreciated.

We have 4 ESXi farms each with it’s own iSCSI or Fibre Channel SAN, and a total of around 130 VMs.

Given the above, how should be implement Veeam proxy servers? Would it be advisable to have a virtual proxy on each farm? regardless of the number of VMs within the farm? is there a recommended VM to proxy ratio? Or is it more around rate of data change?

Backup Repository -In terms of storage for backup repository is there any rule of thumb for spec'ing storage? Are local disks i.e. 15K SAS in RAID 10 likely to be a bottleneck vs. a dedicated FC SAN?

Any other tips / recommended reading for planning an evaluation?

Thanks

[dellock6]

Hi, your proxies needs to "see" the underlying datastores where you will read VMDK of the VMs you want to backup. If you are going for virtual proxies, they will rely on the hosting ESXi servers, so check which ESXi see which datastores, and deploy your proxies accordingly.
If every ESXi sees every datastore, every backup proxies will be able to backup every VM. In this way, even a single proxy can backup all the VMs.
From here on, sizing the proxy or deploying multiple proxies depends on the number of cuncurrent jobs your source storage and target storage can bear: at some poinf one of the two storage (the slowest one...) will become your bottleneck, and so adding more cuncurrent jobs will be useless.
I cannot tell you which storage will be the slowest, there are many tech specs determining their speed.
Also, target can become your bottleneck even faster depending on the backup method you choose: forward incremental is the lightest, other put higher load on it.

[Gostev]

For FC SAN farms, you may want to deploy physical proxy for direct SAN access processing. And, I always recommend to have at least 1 virtual proxy per farm to enable hot add restores.

Not just rate of data change, but many other considerations too (such as backup windows, backup proxy performance, backup storage performance). As a result, VM to proxy ratio may vary up to 100 times between environments. I can easily imagine a VM that would require the dedicated backup proxy! On the other hand, hundreds of VMs in our test lab can be easily backed up by a single pretty weak backup proxy...

Regarding the backup repository, your proposed spec looks quite awesome

[Gostev]

Luca beat me

[dellock6]

Anton, need to save your reply for future bragging

[HighLevel]

Thanks Luca and Gostev for the feedback, just comming back to planning our Veeam deployment afer being pulled onto another project for the past few weeks.

To expand a little on our VMware clusters:

LAN:
5 Hosts
105 VMs
Fibre Channel SAN (~5TB provisioned)

DMZ 1:
3 Hosts
32 VMs
iSCSI SAN (~4TB provisioned)

DMZ 2:
3 Hosts
15 VMs
Fibre Channel SAN (~2TB provisioned)

DMZ 3:
3 Hosts
15 VMs
iSCSI SAN (~2TB provisioned)


Proxies
--------
I'm thinking a dedicated physical proxy with direct SAN access for the LAN cluster, together with a virtual proxy running on the cluster for redunancy
Then a single virtual proxy in 'hot add' mode for each of the smaller DMZ clusters?


Backup repository
-----------------
I now plan to utilize a Dell R610 with 16GB RAM and 4x4Gib HBA left over from another project connected to a FC SAN as a repository. Could this also act as a second proxy with direct iSCSI access to DMZ 2? - Are we likely to see mcuh of a throughput improvement for the 32VMs or would it not woth it over the hot add VM proxy?
Also what kind of impact is it likely to have on the server acting as a repository?

Kind Regards

[dellock6]

Proxies: your initial design seems good, only check for backup duration since you will have some TBs to process from the DMZs' datastores, especially DMZ1. But this will also depend on the daily CBT you produce, maybe only the first full backup will have an impact (you can lower the impact by slowly adding VMs to the job day after day).

Repository: yes it can also act as proxy, only do some tests here too to see how much data it can ingest, since it will receive data from at least 4 proxies, start small and then try to add cuncurrecny until you reach the bottleneck on the repository. From here you will basically have to accept that speed, or speed up the repository, or add a second one.

Luca.

[Novell2]

Hi,
for a new vSphere 5.1 FC-SAN environment with 60VM's What is the best design? Veeam B&R6.5 physical or virtuelle? If Physical with direct SAN Attached fiber channel? I did not found on the Veeam website pros and cons.