Replication or backup over WAN?

[mikegodwin]

I've noticed my offsite replication jobs seem to transfer less data than onsite backup jobs. Is it normal that a replication job will transfer less data than a backup job? Or should both job types transfer the same amount of data over a WAN connection?

[Vitaliy S.]

Hi Mike,

What do you mean transfer less data? Could you please tell me what figures you're comparing?

Thanks!

[mikegodwin]

I have an onsite backup of 4 servers, and also an off-site replication job for those same 4 servers. I'm comparing the "Transferred" field under the job details.

The backup and replication jobs run about two hours apart at night. Here are examples from a couple of days:

Onsite backup: 6.3 GB
Offsite replication: 3.6 GB

Onsite backup: 14.1 GB
Offsite replication: 4.9 GB

I am interested in setting up backups to a remote site over WAN (instead of replication), but it seems like replication uses less data and I'm not sure my offsite job will complete before the start of the next business day if it takes 2-3x as long to complete.

Am I looking at the right numbers? What do you think?

[tsightler]

Do you have the block size set the same on the two jobs (storage optimization tab in advanced options)? What about compression settings? Typically the numbers should be almost exactly the same, but block size and compression can of course impact this.

[mikegodwin]

I don't see where you can set the block size specifically. Are these settings below what you are referring to?

For the on-site backup job, I check the Storage section, Advanced button, Storage tab:

Compression: Optimal
Storage optimizations: Local target

For the off-site replication job, I check the Job Settings section, Advanced button, Traffic tab:

Compression: Optimal
Storage optimizations: WAN target

[tsightler]

Yep, that's it, and that's probably why there's a difference.

Local target = 1MB blocks
WAN target = 256K blocks

This is configuring the smallest block in which Veeam will detect a change. If you are replicating servers with lots of small changes (typically loads like Exchange/SQL) then using a smaller block size will result in significantly less data crossing the wire.

Note that there are other side effects to keep in mind when using WAN target optimization with backup jobs as opposed to replication jobs. I highly recommend keeping your backup jobs less that 2TB (total size of VMs in job) when using WAN target for backup jobs, ideally even smaller. This is because a backup job stores the blocks from all VMs in the job in the same file and dedupes across them, which can lead to a huge hash table and memory usage on the repository, which can cause performance issues as the number of blocks stored in the VBK file gets very large. Using WAN as opposed to Local optimization means there are 4x as many blocks to track. This isn't as much of an issue with replication since each VM is replicated and stored separately.

[mikegodwin]

Thanks, that's very helpful info. One more question:

A Veeam engineer said for replication jobs you have a Veeam proxy on both ends (production and DR sites), but for a backup job you only have a proxy on the production side. Does that mean a replication job will have better performance/throughput with 2 proxies, or does is the repository in the DR site doing some kind of similar functionality that the proxy handles?

[foggy]

or does is the repository in the DR site doing some kind of similar functionality that the proxy handles?

It is. Repository server also has Veeam agent installed (unless it is a CIFS-type repo, which requires proxying server for the agent to be installed on).