Need solution for large 12 TB VM backups

[AdamDuvall]

Looking for options to better configure and or manage backups of large vm's. Currently what we are running into issues with the current solution. We have configured the job to run as a forever forward incremental because the backup time on a synthetic full was taking 3 and 4 days. We need these backups to run every night and with the currently solution if the incremental/change rate is 1 TB in size the merge operation to the full is taking between 40 to 80 hours to merge. Also we have the job scheduled to do a health check and that process as well can run upwards of 80 hours. I do not have all the specifics of the storage but it is running on a windows repository. It is a DL385p G8 attached to a VNX5800 for the target. It has 12 cpu cores and we have the proxy and repo set to 12 tasks on both. It does not seem like when the merge or health check is running that the processes are taxing the server. So I am at a loss as to why those operations are taking as long as they are. I am open to other questions and or help to what I can do to tune that to not run for 40 to 80 hours.

[YoMarK]

Those merge jobs are mostly taxing for the storage. If it's a windows repository, you should Windows resource monitor(or check WMI counters like CurrentDiskQueueLength ) for Disk Activity time. If it's close to 100%, storage is slowing the process down.
We have some large fileserver VM's and the merge jobs were also taking a lot of time(multiple hours - so not a problem for our backup window). A few months ago we switched to Windows 2016 with ReFS(synthetic fulls only) , and now the total process time is much shorter. This is of course no quick fix, as you will need to reformat storage to switch to ReFS.

[skrause]

Do you have a long enough backup window to run a periodic active full? It sounds like, if you were doing synthetic fulls before, that you have enough available storage.

A large VM with that high of a change rate is going to have issues with any kind of synthetic operation. By doing periodic active fulls you can remove the transform operations which will cut your processing time considerably.

If you need to keep running synthetic operations then you will probably need to look into higher performance storage for your backup repository (fast disks in RAID-10 for example). You may also need to beef up your repository hardware.