Tapeless configuration

[stephrf]

Hi all, I'm planning to implement a TSM instance with no tape library in the short term probably medium term.

My Primary and copy pools will be on NetApp 3270 FC disk. Which can do its own sort of defragmentation - so I'm told - I'm new to NetApp.

Does anyone have any thoughts regarding performance on how to set up the pools: devt=disk or devt=file?

Does anyone have any recommendations on maximum size of volume to present to the TSM instance and I guess also on the instance itself?

Is anyone running a similar configuration?

thanks rob

 

[moon-buddy]

I am running esssentially a tapeless setup but instead of NepApp (although we have NetApp as our primary storage devices), I am using Data Domain.

This is definitely a way to go these days if one needs a tapeless environment. Setup devclass=file with volume sizes of 100 GB or 200 GB depending on the sizes of your files to backup. I suggest the lower value to start with and observe performance.

Enable replication on the NetApp devices so you have your copy pool without ever setting one up. This would mean backup to NetApp directly and replicate data to another NetApp device for the copy pool.

 

[stephrf]

Ed, unfortunately we aren't purchasing the replication software for the NetApp so will have to have TSM managing the copy pool aspect.

Is there a reason why you prefer devclass=file over disk? It seems Ontap can reclaim free space on disk.
thanks Rob

 

[moon-buddy]

Ed, unfortunately we aren't purchasing the replication software for the NetApp so will have to have TSM managing the copy pool aspect.

Is there a reason why you prefer devclass=file over disk? It seems Ontap can reclaim free space on disk.
thanks Rob

Either way does not matter. You can use disk or file for the devclass. Reclamation will work either way. I prefer devclass=file so I can control reclamation.

 

[stephrf]

Ed, thanks. Just realised our Ontap reclaim might not work. I've only ever reclaimed file/tape pools in TSM - can you reclaim disk pools?

 

[moon-buddy]

Not disk pools - reclaim is for sequential and as such will work for devlcass=file.

 

[chad_small]

The recommended setup is to use the FILE devclass. I'd rather have one lost 100GB file than a whole storage pool become corrupted due to write issues depending on how you setup your diskpools.

 

[stephrf]

Chad, Ed, thanks for your input. Rob