Veeam LTO Tape Restore Performance

[Matt.Sharpe]

Hi Guys,

I'm wondering if there are any white papers or blogs people are aware of that compare/outline restores speeds for LTO6/7/8 Tape Drives.

Looking for some real world figures and values for some internal research and future customer conversation etc.

Any materials people are aware of OR even any of your tape restore speed experiences from recent times?

Thanks

Matt

[HannesK]

Hello,
you can get nearly maximum speed of the drive if the infrastructure is fast enough. LTO6 around 150MByte/s (160MByte/s theoretical max) etc. for the other generations. LTO8 is often too fast for SMB customers environment, so the speed is lower than net speed possible.

To get more specific numbers from customers, it would be interesting whether you are looking for tape-to-virtual-infrastructure or tape-to-repository speed.

In both cases, I assume that the numbers are relatively hard to compare, because the "Veeam speed" mainly depends on the load of the backup storage and the virtual infrastructure performance. If you try to restore to a backup storage that is doing backup copy jobs or backups at the same time, then speed will be low. You might see some improvements in V11 with unbuffered writes for the restore to repository (I'm assuming that you use proper block storage)

Off-topic: if you are interested in backup speed, there is an interesting discussion for V10 tape-f29/v10-tape-performance-improveme ... 65431.html

Best regards,
Hannes

[Matt.Sharpe]

Hi Hannes,

We have scenarios being asked where customers have a RTO time of 2-3 days and figures of 30TB+ to restore. We're wondering/calculating if this is possible or likely to be recoverable from tape vs Cloud Restore using some real world values and not theoretical. Obviously a number of 150MB/s may be possible but maybe not likely...

[HannesK]

Obviously a number of 150MB/s may be possible but maybe not likely...

Sure, if there is only one LTO6 drive, then 150MB/s is not the average value. That's because how restore works: https://helpcenter.veeam.com/docs/backu ... ml?ver=100

In general I would say 150MB/s is pretty slow (at least for the customer size I usually talk to). Customers have multiple drives and the bottleneck is usually backup storage or the virtual infrastructure. And sure, network is important... with 4 LTO8 drives you might hit limits in processing of a single tape server.

Also the way of restore is important. A Hot-Add proxy might be the fastest for many scenarios (You did not mention whether it's full VM restore or incremental CBT restore to the same location)

Let's see whether someone posts his hardware configuration and restore speeds (and hope it fits to your customer size)

[soncscy]

Hey Matt,

Historically, the tape vendors have shied away from the read performance because "how" you write to tape is so much more important than the maximum potential.

https://www.dynamicsolutions.com/assets ... _Paper.pdf

Dynamic solutions has some articles for LTO 6 and 7, but this is under the most idea of ideal situations. Even the ITDT/HPE read tests aren't relevant because you're not testing the potentially random seek time, you're just testing the max raw read.

Unless the Veeam team knows more, the only way I know to get restore estimates is to do a live test. Get actual values for your environment (the on-tape positioning, the hardware seeking, and the actual restore infrastructure)

The rest is simply inaccurate.

[Matt.Sharpe]

Thanks Harvey! I can understand different environments will 100% be different. I mean the target infrastructure will differ and play a big part in this. When customers have very specific time values for restoration it is always better to get them to test without promising anything. Some 'real-world' values certainly help prove a point at least rather than going of what it "could" achieve but in the end doesn't.

I will certainly give your White Paper link a read though.

Thanks

Matt