Different speed for LTO3 and LTO4

[Puffy]

Recently we have acquired a new library LTO4 drives

To test them I use the same policy to backup on LTO3 and LTO4 drives. Tapes is still using LTO3 tapes for both type of drives.

I found that the speed that I get for LTO4 is at least 4-5 times faster than using LTO3.

I was wondering whether there are some settings to be tweaked in the media server with LTO3 drives.

Thanks.

[sdo]

Speed is always a tricky issue.  You have to rule out SAN paths, SFP light levels, quality of cables, SAN error counts, hops, buffer credits, fabrics... etc... etc...

LTO drives are very highly engineered and can be quite twitchy about which media they're best suited to.

Are there any tape drive firmware updates available?

All LTO drive manufacturers specify a list of recommended media.  Is your media brand on the drive manufacturer's list?  If not, then you have no option but to test a different brand of media.  Maybe your first (cheap) action will be to buy a couple of test media from a different brand on the list supported by your manufacturer.

Also, which media are being written by which drives, maybe you can identify an issue by filling in a table like this?

Drive Brand      Interface   Driver Firmware       Media Brand      Drive+Media Recommended     Speed

HP LTO3          SCSI-3      blah                      HP LTO2           Probably (?)                        40 MB/s

HP LTO3          FC 2Gb/s   blah                      Maxell LTO3      Unknown                             80 MB/s

IBM LTO4         FC 4Gb/s   blah                      Fuji LTO4         Yes (?)                               120 MB/s

etc...

[S_H1]

Hi,

If LTO3 attached to NBU 6.0, and LTO4 attached to NBU 6.5, The difference between 6.0 and 6.5 might influence. The performance is improved by 6.5.

For example, see "NetBackup 6.5 Backup Planning and Performance Tuning Guide  P.125".

-----------------------------------------------------------------------------
The defaults for these values are 10 for child delay and 15 for parent delay.
-----------------------------------------------------------------------------

 6.0: the child delay was 20 msec, and the parent delay was 30 msec.
 6.5: they have been changed to the half.

bptm log are the following.

[6.0 bptm]
--------------------------------------------------------------------
<2> io_init: child delay = 20, parent delay = 30 (milliseconds)
--------------------------------------------------------------------

[6.5 bptm]
--------------------------------------------------------------------
<2> io_init: child delay = 10, parent delay = 15 (milliseconds)
--------------------------------------------------------------------

And, NET_BUFFER_SZ also changed (UNIX only).

The default value for NET_BUFFER_SZ has changed into a setsockopt() since 6.51. But prior to 6.5 it was 32032 bytes. (See ET1045014)