veeam and offsite dr and backups

[tom11011]

Hi group.

I wanted to get an idea of the best way to architect out the use of veeam across a geographical distance for backups and replication. For background, we will have a 20mb/s link between the 2 sites and about 30 virtual machines.

At site A, we are looking at 3 vsphere enterprise servers and a san.
At site B, the same.

What I'm thinking is replicating vm's from site A to site B, then backing up the vm's at site B, all using veeam. Does this type of replication and then backup sound reasonable?

If this sounds reasonable, is backup hardware required? IE- can we install veeam on a virtual machine at site A, and then direct the backups to a virtual machine (cifs share) at site B with lots of disk space? This would give the san at site B something useful to do.

I guess my main questions are as follows:

1.) Is replicating from site A to B, then backing up the vm's at site B reasonable
2.) Is backup server hardware required considering we have a san at site B.
3.) Should veeam be installed at site A or site B for replication and backup.

Thanks for your feedback.

[dellock6]

Hi Tom,
for replication to work at its best, you need a Veeam Backup Server, does not matter in which site, but you then need two veeam proxies, one at each site. They will exchange VM data moving over the line only the delta data.
About the backup hardware, we do not know how many space you need for replica and how many Gb you have in the DR San, so we cannot tell you if the DR san can hold both replicas and backups.

Luca.

[tom11011]

When you say I need a veeam backup server, do you mean I need hardware? If so, for compression?

For reference, the SAN at site B for backups would be at least 4 times larger than san A in order to hold backups and the replicated veeams.

Does that help?

[dellock6]

Hi Tom,
Veeam Backup Server is the central installation of Veeam Backup, it can be a physical machine or a virtual one, there are many considerations to do while choosing which solution, the most common reasons for going physical are the need to do san mode backups in a FC storage or a high cpu power to do processing activities on the backups (high dedup and compression, synthetic backups, transform operations).
From the backup server you command the proxies, and you need one of them at each site. The backup server can also act as a proxy.

It the san at site B has that size, it can be a good solution for doing both backups and replicas.

Luca.

[Vitaliy S.]

1.) Is replicating from site A to B, then backing up the vm's at site B reasonable

Yes, but be aware that VMware CBT will not be used for replica backup jobs.

2.) Is backup server hardware required considering we have a san at site B.
....
For reference, the SAN at site B for backups would be at least 4 times larger than san A in order to hold backups and the replicated veeams.

Since you're going to use your SAN as a destination target for the backup jobs, you can either go virtual or physical. If you prefer to go virtual you can leverage NPIV technology to mount LUN to VM as pRDM disk, format it with NTFS and use as target for your backups, thus making you backups completely LAN-free.

3.) Should veeam be installed at site A or site B for replication and backup.

As Luca said - it doesn't matter. Veeam backup proxy location is what matters most. Please review our sticky F.A.Q. for more information if required.

[tom11011]

what about compression? With a backup server, would compression be offloaded to the backup server? I'm concerned without a physical backup server, will memory and cpu be effected on the esx servers?

Also, if you had 2 sites, an open checkbook, and veeam, how would you architect this out?

Why would CBT not be available when backing up the virtual machines at site B? Can't I just create a normal backup for the vm's at site B? How woould veeam even know they are replicas?

[Vitaliy S.]

what about compression? With a backup server, would compression be offloaded to the backup server? I'm concerned without a physical backup server, will memory and cpu be effected on the esx servers?

Compression is a CPU intensive task, which is performed by our backup proxies. If you have a physical proxy box, then ESX(i) server's performance will not be affected by this. See more info about our components in this thread: >>> READ FIRST : [FAQ] FREQUENTLY ASKED QUESTIONS <<<

Also, if you had 2 sites, an open checkbook, and veeam, how would you architect this out?

For replication I would install backup proxy on a physical server to use a direct SAN access (FC) on the source site. If this is not possible, I would recommend to go with virtual backup proxies and Virtual Appliance (Hot Add) backup mode configured for backup/replication jobs. This should guarantee you the best performance possible.

On the DR site, I would suggest to use virtual backup proxy to populate replica VM disks via Hot Add process from this proxy server.

As to the backup jobs, use the same source site configuration and again...use a dedicated server (physical or virtual) as a backup repository to write your backup files directly to your SAN storage.

Why would CBT not be available when backing up the virtual machines at site B? Can't I just create a normal backup for the vm's at site B? How woould veeam even know they are replicas?

For more details about this, please review this topic: Backup the replicated VMs. Hope this helps!

[tom11011]

Thanks for responding. Some of the terminology you are using I'm not entirely familiar with. If I understand what you are saying, the following would be the ideal setup.

Site A
3 esx vsphere 5 servers
1 san (iscsi in my case)
1 hardware backup server with veeam installed, no storage required

Site B
3 esx vsphere 5 servers
1 san (again iscsi, with 4x the amount of disk space as site A)

I would use veeam (installed at site A backup server) replication to replicate my vm's from my 3 esx servers at Site A to my 3 esx servers at site B.

Then, I would backup my replicated virtual machines at Site B. However, I cannot use CBT on these vm's at site B. I would simply create a large windows virtual machine at site B to house all of my backups. I would like to use reverse incremental. The backups would take a little longer.

Is this the ideal setup?

[Vitaliy S.]

Almost

Make sure you deploy two proxies on the source site and on the target one for better replication jobs performance. Also I would strongly recommend against storing backup files on VMFS volumes (inside large VMs), as it is not a good practice for a number of reasons. If I were you I would just present large NTFS LUN via iSCSI initiator to your backup repository server.

[tom11011]

ok, then could the server that will hold the backups at site B be a virtual server with a D drive that directly mounts the san?

[dellock6]

You can make that VM be at the same time a veeam proxy (to receive the deduped/compressed data from the source proxy) and a backup repository, with a ntfs disk mounted from the san. This disk can be a RDM or a vmdk.

Luca.

[tom11011]

Thanks everyone for your responses. One final question, is there any benefit at all for having a hardware backup server either at Site A or B in this scenario? Does it get me anything I cannot get with just a virtual backup server and proxies?

[dellock6]

Since your san is iscsi, you can use san mode even using virtual proxies without needing hba; another reason for going physical is having some job involving high cpu resources, like reverse incremental, synthetic full or others, but we for example managed to do this anyway with virtual proxies having many of them, all with a fair amount of vcpu, and spreading jobs among them.

Luca.

[Vitaliy S.]

Yes, I would also recommend to go virtual, since you're not using FC SAN.

[tom11011]

>>>another reason for going physical is having some job involving high cpu resources, like reverse incremental, synthetic full or others


Well, we do like reverse incremental alot! I would like to stay with reverse incremental backing up the replicated the vm's. So in this case, where best to put the physical server in order to use the cpu resources of it?

[Vitaliy S.]

Reversed incremental is an IOPs intensive task, so you need to have fast storage disks on the target site. In your case I would recommend to go virtual and configure backup proxy servers according to our system requirements.

[teksyc0m]

I have a similar scenario with Site A and Site B, but I would like to replicate VMs between sites and do local backups. I am thinking of putting a Veeam backup server (virtual) at each site and creating the replication jobs at each destination site, pulling the replication across the WAN. In the event of an outage at one site, the other site would have the replicated VMs and the backup server to recover them.

Is this an ideal setup?

[foggy]

There's no ideal setup, but this one looks very good.

[Claushoumann]

Hello all

We're facing a much similar situation - I am just wondering if there is any way to achieve the same as the other 2 (replicating SAN's and backup of virtual machines locally at both sites using veeam, and then also incorporating a deduplication appliance or software solution?

[Vitaliy S.]

Since compression is performed on proxy servers, you definitely need to have enough CPU power on these servers.