The prediction of the Best Backup Windows

[DataCenterAI]

Is it a good idea to predict the best backup windows with machine learning?
We can monitor the historic network/storage/CPU/RAM usage, and then predict the best time time slots for backup or VM cloning. Based on those predictions, we can also support an elastic backup frequency and make it automatic rather than reply on a dummy scheduler. Thus, the machine learning technologies can mitigate the performance degradation caused by backup, as well as reduce the data loss caused by improper scheduling. In short, make backup intelligent.

[Shestakov]

Hi Jeremy and welcome to the community!
I really like the idea and we are already working on something similar in Veeam ONE.
The only issue is that it`s hard to tell how much time newly created job is going to take. Much easier is to reschedule existing jobs.

[nitramd]

An interesting idea indeed. This might help reduce storage resource contention.

+1 for making backup scheduling a bit more intelligent.

[Shestakov]

By the way, will you allow Veeam to re-schedule jobs for you or you just want to get suggestion?
Thanks!

[DataCenterAI]

Indeed, some people said RPO/RTO is OMG, and so we developed a machine learning engine which can help Veeam accomplish this feature shortly.

[Shestakov]

Have you already developed it?

[DataCenterAI]

Yes. We've done it and licensed it to a couple of backup vendors and some end customers

[Shestakov]

Nice. Do I understand correctly that you have created your own machine learning job scheduler and sell it to backup vendors?
Different companies use different backup logic and show different statistics to use as data for machine learning.

[DataCenterAI]

Not a scheduler. The machine learning engine provides a set of APIs with predictive models and suggestions for integration.

[Shestakov]

Got it. Thanks!