Our NetApp Nearstore VTL will do the following when you initiate import from a physical tape:

 

It will immediately mount  the tape and start importing(depending on available physical drives, it doesn’t pre-empt the write queue to read anything). The tape is immediately seen by the media server as soon as importing starts. Data will be read from the physical tape and laid down to disk inside the VTL.

 

Then, one of two things will happen.

1: If the VTL has time to read the entire contents of the data, it will simply lay it down on disk from start to finish, and unmount the physical tape from the drive, and let the data sit on disk awaiting restore. When you go to restore, it is very quick. Positioning is instantaneous, and read speeds are blazing, 120-130MB/s

2: If you decide to start the restore before the piece of data you’re interested in is read off the tape, the VTL will stop the import, skip ahead on the physical tape to the file marker requested, and data is passed directly to the media server without being staged on to the VTL disk. In this case, the positioning is slower, as it is in fact the physical tape being manipulated, as well as read speeds, will be whatever is coming off the physical tape at a given point. The rest of the data on tape will be read and imported when it is able to do so

 

All of this is transparent to NetBackup.

The import speed is pretty much whatever the physical drive is able to give you, it’s data dependent and we’re seeing approx 70-80 MB/s on LTO-2 tapes, 80-95 MB/s on LTO-3 tapes.

 

This is how NetApp VTLs work, I cannot comment on other products.

 

So, in summary, the time taken to import is a factor, but somewhat mitigated by the “bypass” operations. Another complication is of course the need to do this preparation step in the VTL before NetBackup gets involved.

 

 

 

 

Knut,

I am just curious how significant is the time taken to import data to VTL.

-Shyam