Hi,
I'm relatively new to NetBackup, but we're seeing slow performance, mostly with duplication jobs to our LTO5 tape library.
Here's a summary of our setup:
NetBackup v.7.6.0.3
Appliance: 5220
StorageTek SL150 tape libary (4 LTO5 tape drives)
SLPs: Write to PureDisk (34TB disk shelf) w/ retention of 5 weeks with a duplication job to tape (w/ 1 year retention)
LTO5 settings: Max concurrent drives = 2; Multiplexing Max Streams per drive = 6; Fragment size = 5000MB
PureDisk settings: Max concurrent jobs = 25; Max fragment size = 5000MB; Max I/O streams = 75/volume
Master (the only media server): Max jobs per client = 30; Min size per duplication job = 80GB; Max size per duplication job = 120GB
We've modifed the SIZE_DATA_BUFFERS for both tape and disk to 262144 and NUMBER_DATA_BUFFERS to 256. We've also decreased our max concurrent tapes drives from 4 to 2, which seems to have helped a bit.
On our duplication jobs, we're only seeing around 15-20mb/sec throughput, while catalog backups are running around 80mb-100mb. Our HPUX/Unix clients are backing up around 60-80mb/sec, but our Windows clients are the ones that are suffering.
Log files are showing that there's too much time being wasted on tape positioning, waiting on full buffers and delays in filling those buffers.
It doesn't appear that we can push enough data to the tape drives and they have to constantly wait for more data before writing it to tape.
Example output from a recent job:
10/21/2015 16:33:07 - Info bptm (pid=21989) waited for full buffer 49882 times, delayed 145940 times
10/21/2015 16:33:11 - end reading; read time: 0:05:45
10/21/2015 16:33:11 - Info bptm (pid=21989) setting receive network buffer to 1048576 bytes
10/21/2015 16:33:12 - Info bpdm (pid=22016) completed reading backup image
10/21/2015 16:33:16 - Info bpdm (pid=22016) using 256 data buffers
Are there any settings that we're overlooking that would speed up performance, mainly on the duplication jobs to tape?