Optimize performance in a file to tape job from NAS shares

[Antonio.Proietti]

Hi, I have a Veeam environment based on a Backup Server in a VSphere VM (for lan-free backup) and a physical Tape Server that has FC connectivity to
an Oracle Tape Library. Veeam release is 9.0.0.902 and both Veeam Backup and Tape Server are Windows 2012 R2.
I have an Hitachi NAS and I mount its root file system (UNC Path) on the tape server with a startup script as a Network drive (H:) and all NICs are 10 Gb/s.
Then I created different file to tape backup jobs each one containing some shares, like H:\HomeUsers\*.* in a job, H:\Share1\*.* and H:\Share2\*.* in another job.
We found that this file to tape backup jobs spent very long time working on "build file tree ..." , also some hours.
There is a better way to mapping a CIFS NAS appliance in order to optimize the step of enumerating the file to backup?

I also have another question: on the properties of Media Pool I enabled the option "Enable parallel processing for jobs using this media Pool" specifying that jobs can
use 2 drives simultaneously. But it never happen also if there are free drives on the tape library. This option is intended for use only when more than a job is working on the same Media Pool, or also a single job can use more than a drive? Sorry for the stupid question but I have some experience for Veeam Backup product but I'm a newbie for its tape job management.

Thanks and regards, Antonio

[Shestakov]

Hi Antonio,

There is a better way to mapping a CIFS NAS appliance in order to optimize the step of enumerating the file to backup?

File to tape backups indeed take long if backing up lots of small files. Have you considered using backup to tape job instead?

This option is intended for use only when more than a job is working on the same Media Pool, or also a single job can use more than a drive?

One job can use it if it has several source jobs.
Thanks!

[alanbolte]

File to tape backups indeed take long if backing up lots of small files. Have you considered using backup to tape job instead?

Are you suggesting that he virtualize the NAS, so that he can create a backup job, and then send those backups to tape?

[Shestakov]

I`m curious what kind of files there are.
If OP means backups it makes sense to write them on tape as backups, if those are just other small files it will not help.

[Antonio.Proietti]

Hi and thanks for your reply and for clarifying the multiple drives usage.
Last week-end file-to-tape job made the full backup and this time it used only 52 minutes to build the file tree..
One job contains 1554233 files and folder for a total of 1,1 TB (so we can say that there are many small files) and the job took 22 hours to tranfer this files to tape.
Last night the same job made the incremental backup and find 31184 files to send to tape, but after 9 hours it was only at 25% of this work and I had to stop the job to avoid any impact on production hours. There is should be something wrong if an incremental backup has a longer duration of a full one.
I configured file-to-tape job in order to save space in our Backup Repository (a DataDomain Appliance) but also because I see no way to create backup from a CIFS share source, (that is mapped by a physical host= tape server ), only the possibility to create a file copy job or a tape job.
Any suggestion is welcome.
Regards, Antonio

[rreed]

We found the best we can get for pulling files across the network to tape is to select the tape server as the server in the job, then just enter the UNC path for its source file/folders. Doesn't have to be a drive letter. Works fine and performs about as well as it gets. You might try some controlled testing w/ different block sizes on the tape drive for performance tuning.

[Antonio.Proietti]

Hi,
I followed the steps you suggested and configured directly the UNC path of the share. I have better performance on this file-to-tape job, but it takes too many time. I also opened a case to Veeam Support, but they said it is by design to compare again all the files for the differential job, suggesting to archive the files before sending them to tape. I think I'll try to create more jobs based on subfolder in order to try to parallelize this workload.

[Shestakov]

I think I'll try to create more jobs based on subfolder in order to try to parallelize this workload.

That should help.
You can also use file mask to copy only desired formats.
Thanks!