Hello all,

I'd like to report some results regarding hardware compression and
tapetype definition for this tape autoloader model:
HP StorageWorks DAT 72x6

I'm using linux kernel 2.4.30 on a Proliant ML350 server. The amanda
is 2.4.4p4, mtx is 1.2.17rel, and mt-st is 0.7.

To test whether the compression was on by default I did:

$ time amtapetype -o -c -f /dev/tape
Writing 512 Mbyte   compresseable data:  37 sec
Writing 512 Mbyte uncompresseable data:  200 sec
WARNING: Tape drive has hardware compression enabled
Estimated time to write 2 * 1024 Mbyte: 800 sec = 0 h 13 min
  
real    9m16.011s
user    0m0.460s
sys     0m8.950s

Then I tryed to set compression off with mt compression 0, and then
re-run amtapetype:

$ time amtapetype -o -c -f /dev/tape
Writing 512 Mbyte   compresseable data:  170 sec
Writing 512 Mbyte uncompresseable data:  170 sec
Estimated time to write 2 * 1024 Mbyte: 680 sec = 0 h 11 min
  
real    18m54.167s
user    0m0.590s
sys     0m12.410s

It makes sense; the tape is slower with hardware compression enabled,
probably because it expands amtapetype random data, taking longer to
write it to tape.

Then I unloaded and reloaded the tape, followed by another amtapetype
rerun:

$ mtx unload
Unloading Data Transfer Element into Storage Element 5...done
 
$ mtx load 5
 
$ time amtapetype -o -c -f /dev/tape
Writing 256 Mbyte   compresseable data:  86 sec
Writing 256 Mbyte uncompresseable data:  86 sec
Estimated time to write 2 * 1024 Mbyte: 688 sec = 0 h 11 min
  
real    10m13.636s
user    0m0.370s
sys     0m5.620s

Not sure here whether the compression was off because of the previous
mt command, or if the tape detected a tape with non-compressed data
recorded on it.

Finally I re-run a full-fledged amtapetype run to figure out the
tapetype entry:

$ time amtapetype -f /dev/tape -e 36g -t HP-DAT-72x6
Writing 256 Mbyte   compresseable data:  86 sec
Writing 256 Mbyte uncompresseable data:  86 sec
Estimated time to write 2 * 36864 Mbyte: 24768 sec = 6 h 52 min
wrote 1132416 32Kb blocks in 96 files in 12018 seconds (short write)
wrote 1138314 32Kb blocks in 193 files in 12187 seconds (short write)
define tapetype HP-DAT-72x6 {
    comment "just produced by tapetype prog (hardware compression off)"
    length 35480 mbytes
    filemark 0 kbytes
    speed 3002 kps
}

real    417m41.924s
user    0m11.640s
sys     4m21.480s

(outch! almost a 7h run...)

Don't understand why the filemark is zero. Anyway, the length seems
much more aproximate than the one obtained with amtapetype with
hardware compression (default) -- see below.

Anyway, I'd like to report the following two entries for this DAT
autoloader, corresponding to the compression switched on and off:

define tapetype HP-DAT-72x6-c {
    comment "HP autoloader DAT 72x6 (compression on)"
    # data provided by Rodrigo Ventura <yoda AT isr.ist.utl DOT pt>
    length 31255 mbytes
    filemark 527 kbytes
    speed 1580 kps
}
                                                                                
                        
define tapetype HP-DAT-72x6-nc {
    comment "HP autoloader DAT 72x6 (compression off)"
    # data provided by Rodrigo Ventura <yoda AT isr.ist.utl DOT pt>
    length 35480 mbytes
    filemark 0 kbytes
    speed 3002 kps
}

It's curious the difference among filemark sizes and speeds (almost 2
to 1). Weird... any explanations?

My plan now to convert my tape library is to run amtapetype (write
something + check compression status) after a mt compression 0,
everyday, for each tape due to the next amdump run.

Cheers,

Rodrigo

-- 

*** Rodrigo Martins de Matos Ventura <yoda AT isr.ist.utl DOT pt>
***  Web page: http://www.isr.ist.utl.pt/~yoda
***   Teaching Assistant and PhD Student at ISR:
***    Instituto de Sistemas e Robotica, Polo de Lisboa
***     Instituto Superior Tecnico, Lisboa, PORTUGAL
*** PGP fingerprint = 0119 AD13 9EEE 264A 3F10  31D3 89B3 C6C4 60C6 4585