DLT IV takes longer than DLT III

[Robert Milne]

Hi All,

I have a dilemma, when running our midday backup using the usual DLT III tapes backup duration is approx 3-3.5 hours in length.

I recently had to swap out a faulty tape for a new HP DLT IV Tape, when using this tape for the EXACT same scheduled backup, during time is significantly increased..
Backup length increases to 5 Hours..

Can anyone explain why this is the case?

My Server, Tape Drive, and Backup Type are below.

HPUX 9000 L1000

HP 8000 DLT Tape Drive

Simple TAR backup

Cheers

Chris

[Shaikh Imran]

Hi,
Lot of questions still..
What is DLT III & DLT IV ?
Is DLT 8000 your current drive ?
What was your earlier drive ?
Assuming that DLT 8000 is the current drive:
Do you have the right media for the right drive ?
I have faced only 2 reasons for DLT 8000 running slow
a) firmware version
b) The head-gear assembly in the drive which i got replaced once successfully.

The second reason (b) is as good as replacing the drive.

Also i will advise you to download LTT and have other check for the drive.
Follow the link:
http://h18006.www1.hp.com/products/storageworks/ltt/specifications.html
download the LTT as per your O.S.
and carry on the test.


Regards,

[Robert Milne]

Sorry..

DLT III and DLT IV are types of DLT Tapes.

I.E. DLT 3 15/30GB and DLT 4 40/80GB

Can hold different capacitys...

I dont think there is anything wrong with the tape drive, (HP8000) as it was recently replaced after a failure.


Chris

[Shaikh Imran]

Hi,
Please check for the SCSI termination.
If possible reseat the terminator.
If you doubt O.S. then pls use other command e.g. cpio and try to get the speed.

BTW,Have you got a fresh brand new DLT drive or it's a replacement (used one). ?
Either case this seems to be a problem with the drive which can only be tested by replacing and testing it.

Regards,

[Robert Milne]

Once again, I dont believe this problem is caused by a faulty DLT drive.

If it were then i would be getting variable backup times everyday..

This is not the case, If i use a DLT III Tape then the backup takes 3 hours, if i use DLT IV then backup takes 5 hours.

The Tape Drive is NEW, and is not defective.

Chris

[Ashwani Kashyap]

I think it might be formating related issue . Have you tried to initialize the DLTIV media on your current drive before taking a backup .

[Robert Milne]

Can you please explain how I can do this?

initialising??

Do you mean a mt stat??

Chris

[Mark Carlson]

Chris,

The DLT-III tape media and DLT-IV tape media are physically different inside the cartridge. DLT-IV has 50% more tape in it which may account for the 50% increase in backup time.

Typically the drive accounts for increased density as part of normal operation. I'm no drive expert but assume there is may be a setup function that would allow maximum speed operation.

Alternately you may be running into a situation where the DLT8000 running with DLT-IV media is outpacing the ability of the your host to stream data. This interesting condition could actually result in a decrease in throughput.

Should be plenty of drive experts out there with experience in your combination/situation...

[Scott McIntosh_2]

DLT 8000 tape drive
with DLTtape IIIXT - 1.5MB/s & 15GB
with DLTtape IV - 6MB/s & 40GB

So the streaming concern is valid. But I'm not sure that it would account for running at less than 25% native rate.

3-3.5 hours at 1.5MB/s is 16GB or so, basically one full tape. Are you sure your backup is taking 3-3.5 hours, or filling up a tape is taking 3-3.5 hours? If the backup, then you're likely not even getting a 1.5MB/s data rate out of tape drive front end (buffer and compression). Which would only become a worse problem with the higher speed DLTtape IV would be wound at.

Thanks,
Scott
HP Support

[Robert Milne]

The backup consists of the backup of files, then a tape readback of contents..

So i imagine that actual writing to the DLT III tape takes only 1.5 hours, and then the read back for the remaining time?

Would i be correct?

Chris

[Scott McIntosh_2]

I don't know. You have to tell us what you are doing.

So now it's not 3-3.5 hours to do a tar -cvf, but to do a tar -cvf followed by a tar -tvf? Or a tar -xvf?

You might want to just download hptapeperf for HP-UX and run that against your tape drive with the two different tapes. I don't find your backup scenario to be a valid test of tape drive baseline capabilities.

It might be true that a DLT 8000 drive is "too good" for your incoming data stream vs. a DLT 4000 drive (the basic specs you are performing at with a DLT IIIXT card), but first determine if you are getting correct performance from the drive first.

http://h10025.www1.hp.com/ewfrf/wc/genericDocument?lc=en&cc=us&docname=lpg50460#N10A0D

Thanks,
Scott
HP Support

[A. Clay Stephenson · Accepted solution]

I think I can explain this. The fundamental problem is that you are using a singly-threaded application (tar,cpio) to feed a fast drive. What you need to realize is that when a DLT is not being fed fast enough to stream, it must write a bit, rewind, and write a bit more -- taking about a 100X performance hit as compared to a streaming device. Even though a slower tape drive may be 4-5X slower than the faster DLT in streaming mode, the very fact that it is slower allows it to stream more and thus avoid the huge 100X performance hit more often than its faster counterpart. This is actually a case of less is more. The solution is to run a multi-threaded backup with multiple reader processes to fed the drive at a faster rate. Fbackup is one possible solution; OmniBack / DataProtector is another.

[Robert Milne]

Thanks for your answers..

Yes we are running a simple TAR backup, the reason for this is because of its portability..

And yes we are running a TAR -cvf followed by a readout of tape TAR - -tvf

Thanks for your response Clay, i think your on the right track. I am aware that FBackup is alot more efficient than TAR, but we would rather not run this.

I think FTIO is a multi threaded backup type, and its backwards compatible with CPIO (Which is nice for portability.

I will try this and let you know the results..

Chris