backup takes longer on LTO drive

[dbase77]

Hi,

I've been running a few tests for our backup on solaris 8 sparc with netbackup businesserver 4.5 GA. The backup took about 7 hours for 93 GB data. Setting as follow:

multiple volume = NO
multiplex = NO
Buffer size = 262144
no. of buffer = 16
network buffer = 262144
tape library with 1 robot and 2 LTO drives.

I want to use ONLY 1 tape drive. What am I missing here? Anybody could share their experience?

Thank you in advance.

regards,
feroz

 

[comtec17]

Are you running a SCSI LTO or Fibre Channel LTO. If you are running a SCSI LTO, what type of SCSI card is it running off of? You also might want to look at your logs to see how many buffers NetBackup is trying to use. I would recommend you changing the buffers to 32.

 

[dbase77]

Hi,

Im running SCSI LTO with ultra SCSI 3. From bptm log file I can see it using 262144 data buffer and network buffer and 16 no. of data buffer.

Before I set it to default which is 32K buffer. Which took about the same time 7 hrs. What I dont understand, If I use both drives and split the filesystems between 2 drives using 2 policy, backup took less than 1 hour. I want to use ONLY 1 drive so I created another policy which contain all the filesystem. This is weird. I should get the same time though.

regards,
feroz

 

[comtec17]

If you want to use one Drive I would suggest you multiplex the backup and split your filesystems into streams. An LTO drive has 8 heads that can write simultainously. This will also increase backup speeds.

 

[dbase77]

HI,

Im new to netbackup. What is the best setting for multiplex? To multistream, I just enable that in policy section right? I suppose I need to allow 1 job per client only as if I allow many jobs then multistream will use 2 tape drives.

I will try this for tonite backup and let you know the outcome.

regards,
feroz

 

[PGPhantom]

93Gb in 7 hours translates to about 3800KB/sec. There are many, many things that can and will affect the speed.

Is this a local backup - i.e. Backup of the server to itself of across the LAN?

If LAN - There are a number of things to check such as speed, duplex, checksum offloading, router, switches, flow control etc.

If local and still slow - Check drivers, device drivers, robotics drivers etc.

I just dealt with an issue where I was getting 1700KB/sec on an LTO drive, I changed the adaptec driver and got up to 20000KB/sec - A huge difference.

 

[dbase77]

HI,

You are right, the speed around 3800KB/sec. This is local backup. No LAN. I know something is not right here.

How do I go about checking the drivers? Im on solaris 8 sparc. Can you give me some guideline, please?

Thank you in advance.

regards,
feroz

 

[dbase77]

HI,

Before I forgot, if I use 2 tape drives at the same time, backup only took less than 1 hour. I used 2 policy and split the filesystem between them.

regards,
feroz

 

[comtec17]

To enable multiplexing it must be set in 3 places.

Storage Unit
Attributes of Class
Schedule of Class

Then in your files list you would create 2 Directives like this:

NEW_STREAM
C:SYSTEM_OBJECTS
NEW_STREAM
D:
This will allow the schedule to use Multiplexing and allow the files to be multistreamed.

 

[dbase77]

HI,

Im on UNIX. How do go about that? If I use multistream, it will use 2 tapes on 2 drives. I want to use ONLY 1 tape and 1 drive.

Plus I have setup bp_start.notify and bpend.notify script. Each stream will use the script ehen it start/stop.

regards,
feroz

 

[dbase77]

Hi,

I got my best result from my backup schedule. Using just 1 tape drive to backup 93 GB data took about 1 hour with multiplex enable to 4 per drive. Speed around 28 MB/s which is very fast compare to my previous test.

Thank you to all of you.

regards,
feroz