StoreEver Tape Storage
1823067 Members
3296 Online
109645 Solutions
New Discussion юеВ

ultrium 448 performance

 
SOLVED
Go to solution
Gabriele Peresson
New Member

ultrium 448 performance

I have an Ultrium 448 installed on a workstation equipped with a
Seagate Cheetah 10K RPM 36.7 GB SCSI disk
and an
Adaptec 29160 LP Ultra 160 SCSI HBA
I have a transfer rate of about 100/110 MB/min
which means 1/2 MB/sec that is 1/30 of the
theoretical transfer rate.
here is the L&TT output, that is somewhat ambiguous.

Final Results

InstallCheck done for HP Ultrium 2-SCSI

Drive Connectivity test passed.

This HBA is supported.
This HBA has a recommended driver installed.
This HBA has a dedicated IRQ channel configured.
This drive is on a dedicated bus.
This drive has a recommended driver installed.

Drive Read/Write test passed.

Drive Performance test passed.

System Performance test passed


Recommendations

The way that the tape drive is connected and configured in your system is seriously constraining its performance.
Check that you have an appropriate HBA at www.hp.com/go/connect and that drivers are properly installed.
See www.hp.com/support/pat for more information.

If you are backing up from this local disk system, then the backup performance will be seriously constrained by the speed at which data can be obtained from the disk(s).
See www.hp.com/support/pat for more information.

Where may be the problem ?
tks in advance
15 REPLIES 15
Lewis Finch
Honored Contributor
Solution

Re: ultrium 448 performance

looks like you have the hard drive and the tape drive both connected to the same HBA. Get a sperate one for the tape drive and that should make a big difference
"You can't lead the orchestra without turning your back to the crowd"
Charles Cabral
New Member

Re: ultrium 448 performance

I also have a Ultrium 448 External Tape Drive. I am also experience crawling speeds when trying to back up. Let me try to explain how we have it set up currently.
I have a Supermicro server with 4 drive bays. Is has built in raid for either 0 or 1 on each of the two drives. I beleive they are serial ata drives. I am using retrospect 7.7 running on this server trying to backup drices on the network. I have tested copying a 500 mb file from another networked computer and it copies in about a minute or so. But when I try to back up a network drive to the tape it performs similar to your speed. I was told by HP that I need to install araid5 card. Does this make sense? I am really not sure how when backing up does it temporarily buffer the data to the dric=ve before it backs up to tape? Is it this that is causing the bottleneck?

Please Help!!

Thanks
MSECO
Frequent Advisor

Re: ultrium 448 performance

Hi Gabriele,

Please check if software compression is enabled in your backup software.
Software compression needs to be disabled because hardware compression is enabled by default.

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=lpg50244


Keep learning!!!
Richard Bickers
Trusted Contributor

Re: ultrium 448 performance

Performance can be affected by many factors but when you're down to 1-2MB/S then something is seriously wrong!

It's not good to put the disks and tape drives on the same bus/HBA as they have to share the bandwidth but that should only halve the transfer rates. Similar for compression - not using compression will limit you to native rates but that's still in the region of 20+MB/S for a 448.

I suspect the HBA/cabling/termination for such low speeds. With SCSI, the HBA and drive negotiate to find out how fast they can talk and if there are SCSI errors found in the process or later then the HBA uses lower and lower speeds to make sure data is not lost.

There's a lot of good advice in www.hp.com/support/pat on how to find the bottleneck with performance issues so I'd recommend reading through that. Use LTT for the performance tests. You can also use www.hp.com/support/lttfaq to help you with the LTT side of things.

You could also post an LTT ticket (*.lzt) to this thread which would allow us (and you) to check the health of the drive, see if there are SCSI transfer (parity) errors reported and also see what kind of transfer rates were actually negotiated.

Richard (LTT team)
It's more interesting when it's gone wrong
Gabriele Peresson
New Member

Re: ultrium 448 performance

Thanks eveybody !,

I solved the problem using a dedicatet
29160 HBA for the ultrium, and now
throughput is about 10MB/sec.

Do you think that an ultra 320 HBA
could significantly increase performance ?


Gabriele
Gabriele Peresson
New Member

Re: ultrium 448 performance

for charles,

i don't understand your question,
you mean that backup throuput is different
if you copy from local disks than if you copy from network disks?

is this correct?
Richard Bickers
Trusted Contributor

Re: ultrium 448 performance

Glad things are going better now.

To answer your final question, 10MB/S is still slower than the tape drive can go so the bottleneck is most likely your disk drives (which you can verify with the performance measurement tools in LTT).

In this case, changing to a faster HBA won't increase your throughput to the tape drive.

Are your disks on the server or across a LAN? 10MB/S would suggest LAN which will hold you up unless you've got Gb speeds.
It's more interesting when it's gone wrong
Gabriele Peresson
New Member

Re: ultrium 448 performance

I did the test on a Windows 2000 server
with 1 raid 5 array attached to an adaptec 2110S ultra 3 HBA
and 1 Maxtor DiamondMax 200GB ATA 133 7200rpm
and 1 Maxtor DiamondMax 250GB ATA 133 7200rpm

the average throughput was 10MB/s,
but tests with L&TT gave me a
performance of 47MB/s with 2:1 compression
and 81MB/s with 3:1 compression.

Richard Bickers
Trusted Contributor

Re: ultrium 448 performance

Those test results look like the device performance tests which focus on the drive (and the figures look pretty good). If you run the system performance test (backup pre-test) you can measure the performance of your disks as an application would see them.

That would give 10MB/S to be consistent.

If your disks go faster than that then you may have an application configuration issue.

I didn't catch whether you'd run the system performance test (rather than the device one).
It's more interesting when it's gone wrong
Charles Cabral
New Member

Re: ultrium 448 performance

Richard,

I am struggling to get our backup server and tape drive setup. We have a Supermicro 7033A-T
http://www.supermicro.com/products/system/tower/7033/SYS-7033A-T.cfm

With this we have installed 3 Seagate Barracuda 7200.9 - hard drives - 160 GB - SATA-300
Connected to a Supermicro storage controller (RAID) - PCI 64 for Raid 5 to the three drives.

The tape drive is a Ultrium 448 connected to a Adaptec SCSI Card 29160 - storage controller - Ultra160 SCSI - PCI 64

We are also using Retrospect 7.5.

We currently getting a transfer speed of 140mb/s trying to save a file from the local drive to the backup tape drive connected to this local drive.

What could possibly creating this bottleneck?
We can not think of anything else.

Please, any help is appreciated, I need it.
MSECO
Frequent Advisor

Re: ultrium 448 performance

Your Ultrium 448 works at 24MB/s native
48MB/s compressed

You can use U3 (Ultra160) HBA or U320 HBA, but is doesn├В┬┤t make any difference.
Backing up from your network (LAN) drives will decrease your tape performance.

Check this out as well:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&taskId=110&prodSeriesId=63910&prodTypeId=12169&objectID=lpg70064

PAT http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=lpg50460
Keep learning!!!
Richard Bickers
Trusted Contributor

Re: ultrium 448 performance

Charles. As MSECO suggests, all our performance advice is on the web at the links he's referencing. www.hp.com/support/pat is just a quick link to the same document.

Your system sounds pretty good with both disks and tape directly connected to the server using different HBAs so it ought to be feasible to get reasonable transfer rates.

Have you run the performance tools in LTT? That is how to find where and what the bottleneck is. Run the system performance backup-pretest on the disks to find out how fast data can be sourced and run the device performance test on the tape drive to check that's working fine (as Gabriele did). It would be interesting if you posted the figures you get.

I'd expect the disks to be the bottleneck and your setup able to source data around the 20MB/S mark but that'll be lower if you have a lot of small files. LTT will tell you exactly what to expect.

In your post you report 140MB/S which I don't understand. That's either way too high or you didn't put the decimal point in - in which case that's not far off what I'd expect.

If you want a disk system to source file data at 30-50MB/S you'll need something like 8 disks in a RAID. This is obviously expensive territory but that's what you're paying for when you buy big disk systems.

At least run the performance tests and let us know what you find. I'm fairly confident it'll be the disks that are the bottleneck.

Good luck.
It's more interesting when it's gone wrong
Charles Cabral
New Member

Re: ultrium 448 performance

Gabriel,

It appears that we have resolved our problem. We had to reload the correct Adaptec SCSI driver. We were gertting a transfer rate of about 2.2mb/s write and 1.9 mb/s read with the PAT test. Now it is 18.2 mb/s write and 15.4 mb/s read, much better.
We are considering using compression to increase the capacity of the tapes.
Will this slow down the backup speed or transfer speed? Also how to we set it up to compress? And how does this effect the retreival process, does it slow this down??

Richard Bickers
Trusted Contributor

Re: ultrium 448 performance

Charles, you should always use the hardware compression in the tape drives. The drives default to using hardware compression but that can be overridden by the backup application. You'll have to check the configuration of the app for this.

With hardware compression you get faster transfer rates for both backups and restores as well as higher capacity (as long as your system can keep up) and it's completely transparent to the backup app.

If you want to see what compression ratios you are getting you can use L&TT to pull a ticket and look under the Drive->Performance section. Do this immediately after your backup.

The ticket also shows whether hardware compression is enabled or not. That's under Drive->Configuration.

See http://www.hp.com/support/lttfaq to see how to do this.

We strongly advise not to use software compression as it will be very slow.
It's more interesting when it's gone wrong
Gabriele Peresson
New Member

Re: ultrium 448 performance

Hello everybody.

I think I've found THE solution at last.

After a couple of tests on different workstations/servers, with different pci/scsi configurations, i discovered an interesting thing (at least in my opinion):

- If you connect on your computer 2 Adaptec 29160 scsi hba, one connected to hard discs (or raid array), and the other connected to your tape unit,
the overall performance is exactly THE SAME as if you've connected disks and tape to the same scsi hba (about 1-2 MB/s). I think that this is due to the fact that both hbas share the same software driver...
- To obtain goot performance (40 MB/s) you MUST connect your tape to a scsi hba that is DIFFERENT from the one connected to disks (I used an AHA-2940U2W).

Hope that this can be of help.

Gabriele Peresson