Veeam Backup & Replicator 6 architecture question

[tedsteenvoorden]

We currently have a couple of physical Backup & Replicator 5.0.2 servers. These servers backup around 15 TB of virtual machines. Each backup server has direct attached storage to store the backup files and is connected through fibrechannel to our SAN infrastructure. Virtual machines are backupped through direct SAN access. We are currently running out of our backup window and an additional backup server has to be implemented.

We are currently evaluating the use of Backup & Replicator 6 and have two implementation scenarios:

Scenario 1: Each physical server gets the Backup Proxy and Backup Repository role. In this scenario each physical server is responsible for the backup of a set of virtual machines through direct SAN access and stores the backups on local storage.

Advantages:
- There is no impact on the virtualization infrastructure

Disadvantages:
- There is no optimal route for backup traffic. Each backup server processes a set of VM’s and these VM’s are stuck to the backup server.
- Scales less than scenario 2

Scenario 2: Each physical server gets the Backup Repository role. In this scenario each physical server gets a network connection for backup data (probably 10 Gbit). Multiple virtual machines will be implemented with the backup proxy role. The backup proxies backup the virtual machines through the virtual appliance mode and will send the backup data to the backup repository servers. Each ESX host has a dedicated 1 Gbit link for backup traffic.

Advantages:
- An optimal route for backup traffic is automatically selected which enables efficient data transfer.
- The scenario is easier to scale by increasing the number of backup proxies.

Disadvantages:
- No use of direct SAN access
- Has impact on the virtualization infrastructure.

These scenarios are based on the Veeam documentation but I am looking for some advice based on real-world implementations. Which scenario is preferred and why? I am looking forward through to your responses.

[Vitaliy S.]

Scenario 1: Each physical server gets the Backup Proxy and Backup Repository role. In this scenario each physical server is responsible for the backup of a set of virtual machines through direct SAN access and stores the backups on local storage.

Advantages:
- There is no impact on the virtualization infrastructure

Another advantage is that you will be using the best possible backup mode due to a direct SAN connection. Also please note that this approach will be completely LAN "free", as your backups will be written to local disks directly.

Scenario 2: Each physical server gets the Backup Repository role. In this scenario each physical server gets a network connection for backup data (probably 10 Gbit). Multiple virtual machines will be implemented with the backup proxy role. The backup proxies backup the virtual machines through the virtual appliance mode and will send the backup data to the backup repository servers. Each ESX host has a dedicated 1 Gbit link for backup traffic.

Why don't you want to combine these two approaches? If I were you I would have left the physical backup proxy in place and would have created virtual backup proxies with VA mode specified for better performance and scalability. In addition to saving backups locally I would consider using some offsite locations too.

By the way, there are lots of existing topics on different backup scenarios and deployments, please use our forum's search to get more details. Hope this helps!

[dellock6]

I will go for scenario 2: the impact on the infrastructure is there also in direct SAN, since the production storage is running other VMs while you are extracting data from it, but is limited by the fact you run backups in non production hours.
Cleaned the path from "disadvantages", you have all the advantages yourself listed. Deploying several VM for proxies can distribute the load on the virtual infrastructure, maybe using DRS rules (if you have it) to keep those VMs on different hosts.