Advanced Deployment on all Flash Storage

[Hoegimator]

Good day

We are running our current Veeam Availability (B&R / VeeamOne) Infrastructure hardware since more then 5 years and it is about time to replace it with a new Advanced Deployment design. We are having close to 200 VM's running on a Cisco Hyperflex All Flash Storage with vSphere 6.7.

Our Advanced Deployment is planned like that:

1x VM for B&R
1x VM for Enterprise Manager, VeeamONE and SQL-Server for all databases
1x VM for WAN-Accel, Mount-Server and SoBRepository with iSCSI access to the Primary and Secondary backup storage
4x VMs for Backup and Guest-Interaction Proxies
1x Physical server for Tape-Server and fiber-channel access to two libraries (5 FC-Tape drives)

Does this ensure load balancing for best permormance or is it better to distribute the different roles to more VMs?
Is it ok to have the SQL server on same VM with VeeamOne and EM?

For a better understanding, I made a visio -> https://1drv.ms/u/s!AqWzGDhvaCPrgvF2eFLTbiTHmSAK_Q

I'm thankfull for any input.

Thanks
Joerg

[HannesK]

Hello,
I would go with the KISS principle (keep it simple) as it is a pretty simple environment.

- all-in-one (including SQL to avoid chicken-egg issues) VBR installation on the physical server does everything (no domain. just standalone)
- if you really use the self-service portal of enterprise manager: yes, a domain-joined VM is a good idea
- one VM for Veeam ONE

ONE and EM can share the db from my point of view, as both are not critical. Yes, you can load balance, but in the end, I'm not sure whether it's worth the complexity.

Best regards,
Hannes

[Hoegimator]

I have to give more information about what we plan to do.

-> Except 1 old physical server we use to connect to our FC tapelibrary, we are planning to have everything on a VMs (no VBR installation on a physical server anymore)
-> The WAN-Accelerator is quiet busy because we have copyjobs from our subsidiaries to the HQ
-> We already have an "Advanced Deployment" with 2 physical servers and several VM's and they are all under heavy load (CPU/RAM)

We will consider the KISS priciple, but I'm not sure if that is the way to go for us.

But thanks for your input.
Joerg