NetBackup DC 4.5 running slow

[shieh]

I have a SUN 280R with 1x900MHz CPU, 1GB memory; connected to Qualstar Tape Library with two SDLT220 via 2GB Emulex HBA and 2GB Fibre Option on the library. Yet I can only reach around 5.5MB/Sec when I do a ufsdump from backup server to tape which is only half speed that Quantum marketed. Qualstar suggested me to install NetBackup to see the differences. So I went ahead installed NetBackup DC V4.5 but only to find it even slower. Are there any parameters that have to set correctly to speed it up?

Thanks ahead for your helps.

Neil

 

[bswip]

What are you backing up? Is it another server over the network, a local backup? NBU will definitely tell you 2 things. One, if you have DNS issues and two, if you have network problems. A third is it will tell you if a client is down.

NBU is also tigtly linked to your hardware so make sure it is configured correctly.

As far as tuning parameters, Veritas' support site has some excellent tips.

http://www.support.veritas.com

We get excellent speeds out of our NBU ---- approaching 12 -15 mb/s

 

[shieh]

There is no network involved in this case. What I tested was "ufsdump 0cf /dev/rmt/1 /" on backup server itself to Qualstar tape library that is directly connected to backup server.

 

[bswip]

You may need to check your /etc/system to see what buffer sizes your using. We have optimized ours at 65536. Here is a portion of my /etc/system for Netbackup tuning:

* Begin tuning info for Netbackup
* Entries are not specific to netbackup
* Entries are for increasing performance in general for high I/O apps

set shmsys:shminfo_shmmin=1
set shmsys:shminfo_shmmni=1024
set shmsys:shminfo_shmseg=256
set shmsys:shminfo_shmmax=134217728

set semsys:seminfo_semume=64
set semsys:seminfo_semmap=64
set semsys:seminfo_semopm=32
set semsys:seminfo_semmni=1024
set semsys:seminfo_semmns=1024
set semsys:seminfo_semmnu=1024
set semsys:seminfo_semmsl=64

set msgsys:msginfo_msgmap=512
set msgsys:msginfo_msgmax=8192
set msgsys:msginfo_msgmnb=65536
set msgsys:msginfo_msgmni=256
set msgsys:msginfo_msgssz=8
set msgsys:msginfo_msgtql=512
set msgsys:msginfo_msgseg=8192

 

[shieh]

The following is what I have in /etc/system according to NBU's suggestion. I copied your setup and run again with ufsdump but get the same perforance. Thanks though.


set msgsys:msginfo_msgmap=500
set msgsys:msginfo_msgmnb=65536
set msgsys:msginfo_msgssz=8
set msgsys:msginfo_msgseg=8192
set msgsys:msginfo_msgtql=500

set semsys:seminfo_semmni=100
set semsys:seminfo_semmns=300
set semsys:seminfo_semmsl=300
set semsys:seminfo_semmnu=600

set shmsys:shminfo_shmmax=734003200
set shmsys:shminfo_shmmin=1
set shmsys:shminfo_shmmni=100
set shmsys:shminfo_shmseg=10

 

[comtec17]

Make sure your Emulex HBA's are not set to Auto Negotiate the Link speed.

 

[shieh]

Did try all Link Speeds (auto/1GB/2GB) that also matched with speed on the fibre setup of Tape Library. I am ended up with set both on auto now. Performance still stays at around 5.5MB/Sec.

 

[85235]

I don't know whether the below setting can help you or not?
touch TWO files
/usr/openv/netbackup/db/config/NUMBER_DATA_BUFFERS
/usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS

content of NUMBER_DATA_BUFFERS is 32
content of SIZE_DATA_BUFFERS is 131068

 

[85235]

Remember to inform you that you need to restart netbackup after create TWO files

 

[shieh]

I created two files with different value because I have DLT tape drive. Do you think it is right or not? I am happy now I can reach 8MB/Sec even it still has room to improve.

/usr/openv/netbackup/db/config/NUMBER_DATA_BUFFERS (126)
/usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS (65536)

By the way, I also touch /usr/openv/netbackup/NET_BUFFER_SZ with 65536 to match with SIZE_DATA_BUFFERS.

Thanks for helps from all responses.

 

[85235]

Hi, shieh
I have some experience to set the different value and get the following result

NUMBER_DATA_BUFFERS / SIZE_DATA_BUFFERS / Result (in MB)
8 / 131072 / backup error with status code 11
16 / 128 / 0.3
4 / 129024 / 10
16 / 129024 / backup error wirh status code 11
8 / 129024 / 8
2 / 129024 / 4
4 / 131027 / 5
32 / 131068 / 15

Pls share your setting if you have better value
Thanks..!!!

Rgds
85235

 

[talshekel]

Buffer settings must be set purely on what kind of data you are backing up and what kind of hardware you are using to back it up.
First contact the hardware vendor and find out what the buffer setting their hardware is optimized for.
Use that as a a foundation.
If you are backing up large file set a larger buffer and maybe less of them. If you are backing up small files set a smaller buffer size but with more buffers running concurrently.
When you get an error 11 it is because you dont have enough RAM for the buffer sizes set and amount of buffers used.
This is a trial an error process.
You wont benefit too much on changing buffer sizes on an NT media server - you will only really benefit from configuring buffer sizes on UNIX.
Anyway - Cheers

 

[shieh]

In my hardware setups(SUN 280R,Qualstar Tape Library, 2 SDLT220, fibre connection), I ended up with SIZE_DATA_BUFFERS = 128090 is the best(10mb/sec) for me now. 128090 is the suggested value from Quantum for SDLT 220. As far as NUMBER_DATA_BUFFERS, it looks like taking whatever I assigned and got better performance if it increased.