How much capacity needed for GRT based backup ?

[cc1702004]

Hello,

I am new to GRT. So far I read GRT based backups need disk. Is there any guideline or documentation that let me calculate how much disk space we need to allocate for a GRT based backup ? For example, if I need to backup some Exchange databases over 10 TB, how much disk space do I need to allocate for backup and restore of those databases ?

Thank you

Chan

[RiaanBadenhorst · Accepted solution]

Database size x number of copies you want to be able to restore from (retention period)

[Marianne · Accepted solution]

It depends on whether you will be using basic disk or dedupe disk (MSDP).
The second important point is if you are going to stage to tape or some other secondary storage for long-term retention.

If we assume basic disk with staging, always provide a little bit more than actual database size. About 12 TB for 1 copy.

Because you can only restore GRT from disk, ensure you have enough capacity for 2 full backups.

My advice - 25TB or more.

I am sure other experts will be along soon! 
Let's see what they advise from personal experience.

[Nicolai · Accepted solution]

GRT backup does not uses more/less space than non-grt backup.

The GRT requirement for disk based backup is based on how data is read, GRT can have a random read pattern that disk can fulfill, but tape can't.

[Lowell_Palecek]

Exchange GRT backup does add to the size of the image. During the GRT phase of the backup, bpbkar32 brings up the databases from the backup image, using the magic of nbfsd to present the image to the JET database engine as writeable storage as the transaction logs are rolled forward. Once the databases have been brought up, bpbkar32 writes a "delta" file of the change in the .edb files.

The size of the delta file depends on how many uncommitted transaction log files you have at the time of the backup. If you have a few dozen uncommitted log files, the delta file may be small in comparison to your multi-terabyte backup. I've seen customer cases with 10's of thousands of log files, but there weren't many uncommitted files. You can use eseutil to tell where your checkpoint is and what your log generation point is. The difference is how many log files have to be rolled forward during GRT processing.

[cc1702004]

Thank you for your valuable feedback