08-17-2013 07:04 AM
Hey friends, I could use a sanity check here. Goal is to improve backup performance of a DAG cluster node, as (to me, and I'm fairly new to DAG backups) the number of buffer fills and delays is excessive (see below.) Backups are completing successfully ... just slow (i.e. 5-6 MB/sec, written to a PureDisk STU on the media server.) I understand that slow perf is linked to high #s of fills and delays, so I understand why the slow write rate is there ...
Environment is as follows: NetBackup 5230 Appliance (master and media servers) running 7.5.0.6, and Windows 2008R2 running Exchange 2010SP2 (7.5.0.5.) There are approximately 3000 mailboxes associated with this particular database.
On the media server we've configured
* NUMBER_DATA_BUFFERS_DISK to 512
* SIZE_DATA_BUFFERS_DISK to 1048576
On the Exchange clients (Win2008) we've configured
* communication buffer size to 256K
...per the various technotes, other postings here, etc.
Solved! Go to Solution.
08-18-2013 03:44 AM
You can try reducing size_data_buffers to 512/256 and check if it helps (i knw sounds bit odd, but truely performance tuning varies from environment to environment)
Never saw this high values in number_data_buffers (512 !!!!)
try it with 16 or 32
Refer performance tuning guide for 7.1 (applicable for 7.5 as well)
Also have a look at http://www.mass.dk/netbackup/guides/49-netbackup-buffer-tuning.html
08-18-2013 03:44 AM
You can try reducing size_data_buffers to 512/256 and check if it helps (i knw sounds bit odd, but truely performance tuning varies from environment to environment)
Never saw this high values in number_data_buffers (512 !!!!)
try it with 16 or 32
Refer performance tuning guide for 7.1 (applicable for 7.5 as well)
Also have a look at http://www.mass.dk/netbackup/guides/49-netbackup-buffer-tuning.html
08-18-2013 05:47 AM
Thanks for the reply! Your comment holds true still where the target device is DISK, correct? (in other words, the # of buffers is being specified in NUMBER_DATA_BUFFERS_DISK.)
Not doubting your suggestion, just confirming. Again, thanks.
08-18-2013 06:34 AM
yes. your assessment is correct :)