Comprehensive data protection for all workloads
Post Reply
vash
Novice
Posts: 9
Liked: never
Joined: Jan 27, 2011 10:08 pm
Full Name: ash
Contact:

Design recommendations

Post by vash » Jan 27, 2011 10:22 pm

I’m looking for some design recommendations and feedback on backup strategy for my new Veeam B&R deployment. The environment is quite large with approximately 850 VMs to backup so I would like to hear what other admins are doing to properly backup large scale environment.

Here is a breakdown of the environment:
80 hosts with ESX4 and ESXi 4.1 in 30 clusters all in 1 vCenter. Hosts will be migrated to ESXi 4.1 soon
Some clusters are located in remote locations connected by WAN links
Approx 850 VMs. Windows/Linux mix
Combined used space is about 50 TB
Storage is FC SAN and iSCSI. Storage is shared between hosts within same cluster for the most part.
Traditional agent-based backup is used to backup the VMs nightly. Since this is managed by a different group so I don’t foresee it going away even after Veeam is deployed.

My preliminary design looks like this:
Veeam servers will be hybrid using both physical and VMs- -
2 Veeam physical servers for the 2 largest clusters using Direct SAN mode.
Veeam servers deployed in VMs using Virtual Appliance mode for the smaller clusters such as the ones in remote locations
Use Veeam to backup OS drives only. Data drives will be backed up using the traditional backup agent. ( I know I’m not taking advantage of all the cool features, but it’s kind of a political thing) :roll:

Questions:
I do like the ease of using a VM for the Veeam server, but I’m seeing faster backup speed using SAN mode on a physical. Is this what you guys are seeing?

Does it put more stress on the storage array when using the virtual appliance mode with the disk mounts/dismounts?

This is probably an “it depends” question, but what’s the typical dedupe/compressed ratio?

Will the restore of a whole VM be transactionally inconsistent because it’s backed up by two different tools? I’m not concerned about Active Directory or any databases restores because other 3rd party recovery tool is used.

I appreicate any thoughts/comments on this!

Vitaliy S.
Product Manager
Posts: 23006
Liked: 1560 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Design recommendations

Post by Vitaliy S. » Jan 28, 2011 3:49 pm

Hello,
vash wrote:I do like the ease of using a VM for the Veeam server, but I’m seeing faster backup speed using SAN mode on a physical. Is this what you guys are seeing?
What is the current CPU load on your virtual backup server? Have you specified 4 vCPU for that very VM?
vash wrote:Does it put more stress on the storage array when using the virtual appliance mode with the disk mounts/dismounts?
I don't think so, as a matter of fact, I believe you shouldn't have any signifficant differences in I/O performance rates for both modes.
vash wrote:This is probably an “it depends” question, but what’s the typical dedupe/compressed ratio?
Yes, it depends :) You may search these forums for some examples from our existing customers. Hopefully, one could jump in to this thread and share his numbers.
vash wrote:Will the restore of a whole VM be transactionally inconsistent because it’s backed up by two different tools? I’m not concerned about Active Directory or any databases restores because other 3rd party recovery tool is used.
Yes, if you have application-aware processing enabled in Veeam job settings.

Thanks!

vash
Novice
Posts: 9
Liked: never
Joined: Jan 27, 2011 10:08 pm
Full Name: ash
Contact:

Re: Design recommendations

Post by vash » Jan 29, 2011 6:09 am

The Veeam vm has 4 vCPU and 3 gb of memory assigned to it. I tested backing up the same VM using SAN mode on a physical veeam server then virtual appliance mode on a VM and consistently saw better results on the physical. During the initial run of a 12 GB VM, it backed up at 45 mb/s on the physical and 35 m/s on the VM. I'd much rather use a VM so let me know if there's anything else I can tune to get better results.

Vitaliy S.
Product Manager
Posts: 23006
Liked: 1560 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Design recommendations

Post by Vitaliy S. » Jan 30, 2011 11:02 am

Though these modes should have almost the same peformance results, I should admit that some difference might be expected. By the way, do you use the same destination storage to save your backup files in both jobs?

If you want to stick to a VM mode, then with CBT enabled I believe you can do that while being able to fit in to your backup window.

vash
Novice
Posts: 9
Liked: never
Joined: Jan 27, 2011 10:08 pm
Full Name: ash
Contact:

Re: Design recommendations

Post by vash » Jan 30, 2011 9:25 pm

The destination datastores are different. The physical is using iscsi storage that's mounted to it whereas the VM is using storage on a VMFS datastore on FC SAN.

The problem with enabling CBT is that the environment still has a alot of VMs using hardware version 4. There's no plan to upgrade to version 7....at least not all.

Your comment on enabling application-aware processing from your first post....Did you mean I should enable it to ensure backup is transactionally consistent?

Vitaliy S.
Product Manager
Posts: 23006
Liked: 1560 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Design recommendations

Post by Vitaliy S. » Jan 30, 2011 11:40 pm

I guess the reason why you have different processing rates in both modes is different destination storage. Try choosing the same target to compare all the rates.

Yes, to have transactionally consistent backups you need to enable application-aware processing option.

Post Reply

Who is online

Users browsing this forum: Alexander S., Artem Filatov, christopher-swe, FSenturk, kelaros, SHL and 52 guests