DLT IV TAPES - PERFORMANCE

[Craig Clark_1]

I am currently running an HP 80e (DLT8000) ape drive on a Windows 2000 system with Arcserve 2000.

The issue I have is that the unit appears to get very fussy over individual tape batches.

Currently we have had to return 10 Maxell tapes all from the same batch because the unit speed droped from 330MB/min to 40MB/min over the length of the tape (normally after 500MB the rate starts dropping).

Now I know (or at least think) the unit is fine because we purchased 20 Maxell tapes at the same time and onlt 10 tapes exhibit the problem.

When the problem was first noticed it appeared to be random however after logging the issue closely it was noticed that this only happened on specific tapes, we then noticed from the stamp on the bottom of the tape that all those effected came from the same batch. Problem solved - well not quite.

I have now obtained some brand new HP tapes and once again the unit only likes certain tapes from those purchased. The other less fortunate tapes are snubbed by the system. Now I know HP tape warranty to cover tapes for 20 years but I'm sure they will claims that the tapes have no problem - except for the fact that my system will only backup at 330MB/min on the tapes it likes and anything else it will struggle at 90MB/min.

The question I have is there any benchmark utility out there that will chart the backup performance of a tape over its entire length.
HP Tape Tool is simply not sophisticated enough.

NB All tapes are brand new and only used on this unit. I am aware of the issues regarding DLT8000 and VS80.

[Alexander M. Ermes]

Hi Craig.
1) Try to download the ltt tools and test your environment.

2) what about the file sizes ? if the files are small, the tapedrive will not go into streaming mode ( repositioning after every file takes a lot of time ).

3) Do you work with software compression on ?
Try to switch it off, because the hardware compress is much more effective and does not work well together with software compress.

I hope, that this helps.
Rgds
Alexander M. Ermes

[Alexander M. Ermes]

Hi Craig.
Here the URL for the ltt tools :

http://h20000.www2.hp.com/bizsupport/TechSupport/DriverDownload.jsp?pnameOID=406731&locale=en_US&taskId=135&prodSeriesId=406729&prodTypeId=12169

Rgds
Alexander M. Ermes

[Craig Clark_1]

LTT tools are useless for the problem described above. They are far to blunt.

I have run LTT diagnostics against the system the drive is upto date and a cant see anything that stands out.

The file sizes are mixed and as described above I have a dedicated TESTJOB that backups up a specific VOLUME full of various sized files.

Tapes that the system like complete the backup in approx 3mins at an av rate of 330MB/min.

Tapes that the system does not like complete the backup anything upto 20mins with av. rates as low as 50MB/min.

The above is repeatable and I can predict the results for tapes run through the test once (they always behave the same no matter what time of day or what else is being run on the system). All the tapes being tested currently are brand new shrink wrapped tapes.

From observation the tapes that do not perform well all appear to start well at 330MB/min but by about 50% through the job (some 600MB into the job) the backup rate starts to drop. Yes compression is switched on. In order to isolate the problem further I may change the directory to contain a commonly available file such as the Microsoft Service Packs. These should be sufficent in size to allow the drive to max out (however in the test above I would concider an average of 300MB/min and above max speed.

As the original post what I am looking for in an application the will produce data rates across the entire tape lenght so I can produce a graph to show DataRate vs TapePosition. The perfect graph should show a straight line. From the observation above the graph slides down for the 'unliked' tapes. The alternative is to sit and document throughput rates by hand during the testjob!

The only explination I can think of at the moment is the the drive starts to lose the tracking track on the tape. Maybe the drive is drifting from true alignment which means marginal tapes are effected as above whilst very good tapes are okay.

My only option is to return tapes as faulty to the manufacturer as the system above demonstrates that the results are repeatable and that tapes of the same brand are accepted without problems.

PS Just for the record I have used LTT to do Write and Verfy tests to the tape however these test are limited to 2GB data sizes maximum and the results are not consistent.

[Craig Clark_1]

One thing I do need to check is the block size currently selected via Arcserve however this still does not explain why some tapes are fine straight out the wrappers whilst others are useless.

I guess ultimately it could be as simple as the handling and storage whilst in the sales channel!

This could have explained the first variance however using different suppliers and brands the same is true. Or maybe for the first time I am scrutinising every new tape before it goes into production.

[Never had these problems with DAT or SLR]

[Craig Clark_1]

PS Just for the record I have used LTT to do Write and Verfy tests to the tape however these test are limited to 2GB data sizes maximum and the results are not consistent.

The above comment related to the fact that all you get in the results is passed and a time for the test section. These are meaning less as it does not give any indication of data rate across the media (only and average if you use some maths).

[David Ruska]

Craig,

If you can run the backup that is slow on an individual tape, and then run HP LTT and generate a support ticket WHILE THE TAPE IS LOADED, I can look at the read/write error counters and see what type of error rate you are getting. If performance of one data set varies dramatically between tapes (and consistantly), usually it's an error rate issue. The device analysis section of the support ticket may also print out the number of corrected errors.

The reason you need to do this while the tape is loaded is due to the DLT drive clearing its error rate counters as soon as the tape is unloaded.

Also, I'll mention that the DLT drive in particular does not like small block sizes and can easily fall out of streaming which slows things down considerably. You should use a block size of 64K or better.

HP Engineering

[Craig Clark_1]

Hi David

Thanks for the offer. When I get into the office I will produce a support ticket (however last nights backup was on tapes known to be good) I will then run a complete job on a known poor tape tonight and upload that in the morning!

I will also grab the external drive and put it on another system for more analysis.

Since the above posts I have used LTT a little more and found that the product support upto 16GB tests (my error). However running these test even on the poor tapes produces aceptable results that are not shown in the real world under Arcserve.

Looking at the tape statistics under Arcserve I can see the block size on the tape is set to 65536

I would greatly appreciate your time to look at the ticket info when posted.

(I probally post a zip of some additional information and multiple tickets all clearly labeled as I could do a short job of say 1GB mixed data that shows the errors).

[David Ruska]

Craig, thanks for the additional information. You can email the support ticket and any other info you have to ltt_team@hp.com.

[Craig Clark_1]

Hi Dave

Thanks. I have sent an email with a copy of the ticket produced today on a 'poor' tape.

I will email more information over the weeknd when I have had time to test on a different system.