I have been reading promising things about the new "modern storage" in DPM 2016.
Can I expect better performance for large SQL recovery point jobs if I switch from DPM 2012 R2?
I would do a complete reinstall of DPM in a new Windows Server 2016 OS.
I would leave the SQL Server as 2014 sp1 or maybe sp2. And I would prefer to leave the SQL Server running on Windows Server 2012 R2.
Today it can take several hours to run large recovery point jobs - sometimes not finishing before the next days jobs were scheduled to start. Depending on how much load the applications using the databases also puts on the server.
The server houses about 12 SQL Server instances each with one large database (0.5 - 3.0 TB each). All databases use simple recovery model. The databases are divided into two protection groups both with the same goals: 1 daily recovery point using Express Full Backup.
When the SQL Server is under high load and it has to perform recovery point jobs, which need to move a lot of data, it tends to get volsnap error 25 in the server event log. Typically after recovery point jobs have been running for a long time. All running jobs fail when volsnap error 25 occurs.
I have seen shadowstorage in use (vssadmin list shadowstorage) up to about 500 GB during such long jobs.
It seems to me that VSS on the SQL Server simply might be overloaded and I guess that I am hoping that another technology is used in DPM 2016. It looks like it is for HyperV, but I am not sure for SQL Server. Hence this question.