Best offsite backup method

[ritterbruder]

I am setting up an offsite repository for nightly backups. This will be over a WAN link with about a 35/5 pipe as the bottleneck. This will be just a windows 7 machine with a shared folder. Should I be doing incremental, reverse incremental or will replication actually work for me? I have about 15 vm's and most of my local reverse incremental vbr files range from about 5gb to 60gb. Sorry for such an amateur question but I need to make sure they are (relatively) quick and effective. thanks!

[Vitaliy S.]

Actually v6 is optimized to off-site backups out of box, so you can simply backup directly to offsite backup repository. To get better performance I would suggest using default backup mode which is forward incremental.

If you're planning to run replication jobs, don't forget to perform initial VM seeding. That would help you a lot since you're planning to do that over the WAN link.

Hope this helps!

[foggy]

Actually, backup mode not really matters from the link perspective but rather affects your target storage. Reverse incremental mode puts 3x I/O load on target comparing to forward incremental mode, but uses much less disk space.

[ritterbruder]

Amazing, Thanks guys! Should I do synthetic fulls offsite as well?

[Vitaliy S.]

Definitely, if you have a remote repository, job's "synthetic full" traffic will be kept within the remote site (no stress on your WAN link).

[ManuG2k]

[merged]

Hi,
I have B&R 6.1.
I need to backup (for disaster recovery) my virtual machines to remote site.
On remote site i have a simple physical windows server 2008 standard.
I don't need replicate my virtual machines, but only emergency backup.

From my site to remote site i have 2Mb/2Mb of connectivity.

What is the best solution for my task ?
On remote server i need only shared folder or i need to install veeam backup proxy ?

Tnx in advance
Manuel

[J1mbo]

How big are your VBK files?

[ManuG2k]

How big are your VBK files?

I have 20 VM's.
Full backup size is big for all VM's (from 1Gb to 60 Gb).
I thought first full backup of all VM's, and later reverse incremental.
But, on destination server (remote site) i need to install Veeam Backup component (veeam backup proxy ?)

[J1mbo]

Just a thought - if this is Hyper-V you could enable the HV role on the remote server then replicate the VMs. If it's VMware you'd need an Essentials license to do the same, which is still low-cost, $500 or so. Could yield better recovery times with the same effort

[ManuG2k]

Just a thought - if this is Hyper-V you could enable the HV role on the remote server then replicate the VMs. If it's VMware you'd need an Essentials license to do the same, which is still low-cost, $500 or so. Could yield better recovery times with the same effort

It's VMware, and have Essentials license.
Then on the remote server should I install Veeam Backup proxy to speed up the transfer?

[Vitaliy S.]

On remote server i need only shared folder or i need to install veeam backup proxy ?

You need to deploy a repository server on your remote site, target proxy is only required for replication jobs.

[ManuG2k]

For remote backup with WAN connection (2M/2M), not replication jobs, normal backup job, the best method for fast backups is incremental or reversed incremental method ?

Manuel

[Vitaliy S.]

It depends on the storage hardware you're going to have at the remote site, but forward incremental backup mode is generally faster because of the less I/O load on the storage disks.

[dellock6]

Just to add to Vitaliy, backup method has an impact on backup storage I/O, but not on the amount of traffic you put on the wan link, all the reverse calculations are done by the remote repository. This is another reason to have one.

[ManuG2k]

Hi,
i try offsite backup.
The backup method is reversed incremental (on storage settings i set best compression level and wan optimization target).

Start time: 9:54:40 AM
End time: 10:19:08 AM
Size: 16.0 GB
Read: 4.0 GB
Transferred: 1002.8 MB
Duration: 0:24:28
Backup size: 992,7 mb
Dedupe: 5.0x
Compression: 3.3x


My connection link is 2M/2M from local to remote site.
The speed for this job is 0,7 Mb/s, or I read wrong the report ?


Manuel

[J1mbo]

1GB in 24.5 mins would be 5.5Mbps, which is clearly impossible. With 3.3x compression it's a more reasonable 1.6Mbps.

[ManuG2k]

1GB in 24.5 mins would be 5.5Mbps, which is clearly impossible. With 3.3x compression it's a more reasonable 1.6Mbps.

how do you get obtain 1.6 mbps from the backup report ?

[J1mbo]

how do you get obtain 1.6 mbps from the backup report ?

I was just using Transferred of 1002.8 MB, Duration of 0:24:28 and compression of 3.3x

So 1002.8MB is approx.8,000 Mb
8,000 / 24.5 minutes / 60 seconds = 5.4 Mbps effective throughput
But it's compressed, so 5.4 Mbps / 3.3 (compression factor) = 1.65 Mbps actual line usage.

I don't know where the de-dupe ratio fits in though as that doesn't seem to follow from any of the other numbers.

If that is the correct way to interpret the numbers, the 1.65Mbps is an understatement of actual WAN usage because of the framing overheads and anything else on the line.

[foggy]

Compression in this case states for the ratio between the actual VM size and the backup file size: with 5.0x dedupe ratio and 3.3x compression you get 992,7 MB backup file from the 16GB VM.

The speed for this job is ~0.7MB/s (1000MB/24m/60s). What is the exact speed of your link? 2MB/s or 2Mb/s?

[ManuG2k]

Compression in this case states for the ratio between the actual VM size and the backup file size: with 5.0x dedupe ratio and 3.3x compression you get 992,7 MB backup file from the 16GB VM.

The speed for this job is ~0.7MB/s (1000MB/24m/60s). What is the exact speed of your link? 2MB/s or 2Mb/s?

2Mb/s

[J1mbo]

which begs the obvious question!

[Gostev]

Should not be using job duration, as there is significant time spent preparing VM for processing, as well as post-processing task. Instead, in calculations like that you should be using actual data transfer time found next to the hard disk line in the job log for this VM.

[J1mbo]

but that would surely be less still?

[Vitaliy S.]

Yes, the duration will be less, however it doesn't seem like 2 Mbit link is used here.

[Gostev]

B&R may not be the only using the link. Alternatively, QoS might be in place there.

[J1mbo]

Gostev, the odd thing is that the apparent transfer rate far exceeds the stated line capacity - even with the job stats.

[Gostev]

Ah, never noticed, sorry about that. We get hundreds of similar support cases every year - most of them are the other way, customers complain they are not getting full speed - all of them are generally due to misunderstanding of actual link capacity, or confusion around bits/bytes. So either that, or kind internet provider, or some bug in the statistics reporting (least likely).

[mayojs]

Link capacity can be calculated (rule of thumb) as 400MB/hour for each 1Mb/s of bandwidth.

That's 90% TCP efficiency, which is reasonable for estimation.

Veeam V6 should be able to deliver this (and possibly a bit more) with multiple TCP sessions and if not (due to latency or packet loss on the bandwidth) you can restore with optimiser such as HyperIP.

2Mb/s circuit should deliver 800MB/hour (0.8GB/hour) throughput
10Mb/s circuit should deliver 4GB/hour
100Mb/s circuit should deliver 40GB/hour

Example: If you have 80GB of compressed payload to move across a 2Mb circuit it will take 80/.8 = 100 hours or around 4 days

There's often a lot of misconception about what performance to expect across links and we try and set expectations accordingly with this rule of thumb.

John Mayo

[danblee]

[merged]

Hello,

For the sake of DR, we'll be implementing a new VEEAM server at an off-site location. The off-site location is on the WAN, but this will be our second server. Our primary VEAAM server is local to the ESXi environment. I was hoping to get some best practices advice regarding how to go about doing this. One obvious solution would be to simply install VEEAM on the new server and do backups across the WAN. Another option would be to just use the drive space on the off-site server and create a new job on the existing VEEAM server, pointing to the off-site drives.

Are there any other methods anyone can suggest? Thanks in advance.

[Vitaliy S.]

And here is another good reading on the "offsite" backup techniques: v6 - How to have a local and off-site backup copy?