After many months, I can finally share with you this new whitepaper I created, and I’m happy it’s finally out because I think and hope it will bring good informations to people planning or designing a Veeam backup repository, especially on the repository performance side.
As I wrote in the intro of the paper, when designing a backup repository for Veeam Backup & Replication, an architect has to evaluate two primary parameters: size and performance. Both values need to be carefully calculated, so that the overall design will be effective for the entire lifecycle of the underlying components, instead of requiring painful forklift upgrades after few years because of incorrect initial sizing.
Estimating the size of a repository is quite easy. Once you know the size of the production environment and the daily change rate, depending on the desired retention period, you can estimate the final size. Also, you can easily retrieve this information using Veeam One, which has dedicated and prebuilt reports for these two parameters, instead of “guestimating” and running the risk of oversizing or even worse, undersizing your repository.
On the other hand, estimating the required performance of a backup repository is more complicated . Obviously, a better solution would be to run some backups “for real” to check the effective performance, but it’s not easy to get access to a demo storage, and most of all, a test backup has an impact on the production infrastructure and its VMs, just like a real backup . Because of this, I worked on developing a testing methodology using an I/O generator and some custom I/O profiles to evaluate the performance of a backup repository without causing an impact on the production environment. Although this methodology is not perfect, and some differences between the results of these tests and a real backup are expected, I think these simulations allow you to have a good estimate. Values are conservative on purpose to eliminate the risk of undersizing your repository.
In the paper, you will learn in high details how each Veeam backup mode works, what kind of I/O profile they create on the storage, pros and cons of each of them. But the paper doesn’t stop at theory: using my lab and FIO (Flexible I/O Tester) I’ve created pre-defined test profiled to simulate each backup mode, and I executed those tests several times, changing parameters like block size and number of threads.
The final result is an interesting table where you will be able to compare the different modes against the same storage, their execution times, and the I/O created. Those tests took a huge amount of time, and they also broke 3 disks in my array 🙂
Now that the paper is out, go and downloaded it here, and let me know what do you think, honest feedback is appreciated, as always.