Skip to main content
  • EN

As the global solutions architect for Veeam's Pure Storage alliance, I have been asked this question many times. I am pleased to report that we now have performance results, and they are exceptional.

But first, a bit on the test environment:

Figure 1
Figure 2

As you can see the Veeam Backup & Replication Backup Proxies were configured as Virtual Machines (VMs).  This was done to reduce the hardware requirements in the lab testing environment.  This also means that the backup traffic to these virtual Backup Proxies is over in-guest iSCSI, which adds overhead to the CPU and memory workload on those proxies.  It is expected that if the Backup Proxies were on physical machines that are physically connected to the storage network, the performance would have been even better.

Performance

Even with virtual Backup Proxies connected via in-guest iSCSI, performance was impressive:

  • Active Full backup performance of 4GB/s (14.4TB/hr)
  • Full restore performance of 4.7GB/s (16.9TB/hr)

And this was with a single FlashArray//C backup target.

For more information, and to see what best practices were used, check out the Veeam and Pure Storage Reference architecture for Veeam Backup & Replication with Pure Storage FlashArray//C

For even more information, this reference architecture aligns with the Veeam Security Blueprint for Pure FlashArray//C with 1000 VMs and 100 physical servers.

Thank you for sharing @MarkPolin 

Nice test and writeup @MarkPolin . Thanks for sharing! #VeeamSpeed 😊

Wow!  Now that is some speed. Always love Pure storage and just more reason for people to invest in it.  😎

Thanks you for sharing, 

Great write up..   I’m about to install 2 IBM FS7300’s and will be doing some serious performance testing of our backups/restores. We also looked at Pure but decisions were made at the top and IBM offered a smoking deal on the boxes. 

I’ve ordered many 25Gb/ 32Gb FC servers as well for proxies/repos and will be doing some testing before putting to production to see the best way to set this up. I’m torn between having 1 repository server or 1 for each SAN.   

Have you tested the performance on a Windows vs Linux repo? I’d assume the Linux server as well?

I’m jealous of that environment. Those X70’s and x90 are beasts. 

Comment


Terms & Conditions