Secondary backup destination to tape design questions.

[kapple]

Hi,

I am redesigning our Veeam infrastructure, and will be introducing a tape library and drive in order to have that 3rd copy of VMs, as well as to have the ability to store backups off-site.

I know that Veeam has an “Unofficial tape library compatibility list”, with both verified and not verified to work products. Beyond actually working with Veeam, I am concerned about both backup speed and capacity of the various tape technologies.

We currently backup 168 VMs. We have 10070 GB of full backups, and 1771 GB of increments. This will increase with our redesign because we will be increasing the retention time for backups. Our source VMs take 39500 GB disk space.

I am considering a Dell PowerVault TL2000 or TL4000, either LTO 6 or LTO 7.

Not having used Veeam with a tape drive, I am not sure how configuring a secondary destination to tape for a backup job will affect our backup window for the initial VM backup to disk, nor the performance of the Veeam B&R server and proxies.

So here are my questions:
1. Does configuring the secondary backup destination affect the initial backup to disk job?
2. If so, does backing up to tape require additional proxies to handle the processing of jobs, so that the primary backup jobs finish within the configured backup window? We currently process all backups between 5:30 PM and 6:00 AM.
3. We run forever incremental backups. For the amount of data we backup up nightly, do you have recommendations on the LTO version, and number of drives to deploy? Since this affects both the write speed to tape, and capacity, selecting both the LTO version, and number of drives, will be critical to assure we design this right the first time.
4. I am not committed to the Dell solution, but this is the hardware I am familiar with, and that can be used to restore legacy backups from our physical servers. Does anyone have a compelling argument for designing a system with one of the other tape drives on the Veeam list?

Thanks in advance for your input.

[veremin]

1.Does configuring the secondary backup destination affect the initial backup to disk job?

Nope.

2.If so, does backing up to tape require additional proxies to handle the processing of jobs, so that the primary backup jobs finish within the configured backup window? We currently process all backups between 5:30 PM and 6:00 AM.

No load will be put on existing proxy servers. The components participating in process are tape server (tape device is connected to) and backup repository (where backups to archive are stored on)

3.We run forever incremental backups. For the amount of data we backup up nightly, do you have recommendations on the LTO version, and number of drives to deploy? Since this affects both the write speed to tape, and capacity, selecting both the LTO version, and number of drives, will be critical to assure we design this right the first time.

It's up to you to decide, but in general the newer LTO the better. The same applies to number of drives - more drives, the better, especially taking into account our tape parallel processing mechanism that we've had since version 9.

Thanks.

[kapple]

Hi Vladimir, and thanks.

Have you seen an issue where where reading from the backup repository during the tape backup job, affects the writes to the repository during other backup to disk jobs? Can the secondary backup job be scheduled to run at a different time, or does it kick off immediately after the initial backup do disk?

Thanks.

[pkelly_sts]

Having big reads AND writes to the same repository can have some impact, depending on how many IOPS the storage has available (e.g. SSD-based storage would probably be fine but I'm guessing you don't have TB's of SSD backup storage).

Tape job can be scheduled outside of main backup job window, it can also be scheduled to start immediately after completion of source backup job.

In your case, running Forever Forward in the source job, the biggest hit on storage will be for the FIRST backup to tape, from there on, it will primarily be just streaming incrementals & when you run a synthetic full (which you will have to at least occasionally when copying Forever Forward jobs to tape) there's still not a huge impact to source repository.

If it helps with numbers, I can easily get a full 2Tb .vbk from disk (which is 12 x SATA disk in RAID-10) to LTO-6 SAS tape inside 4hrs. Providing your storage can deliver throughput, LTO-7 should prove even quicker.

[kapple]

Thanks for the information Paul.

[veremin]

As Paul stated, there will be an impact, though, of considerably low size. Typically, it's primary backup job that kills secondary tape job running long, not vice versa. Thanks.