cancel
Showing results for 
Search instead for 
Did you mean: 

Tuning Network Data Buffers

Colin_North
Level 6
Hi,
I've got one client server that on some mornings is still running its backup, so I have to cancel it. I've checked the BPTM log on my Master Server for "waited for full buffer" entries for this client. One of these entries shows "write_backup_completion_stats: waited for full buffer 14162 times, delayed 14796 times". To me this number seems a bit high, and indicates that there is a bottleneck on the client or network ie. waiting for buffer to fill.

The Communications Buffer Size for this client is 32K, and I had been told on a recent Symantec course that if I got high "waited for full buffer" figures I should REDUCE the size of the buffer. However, from doing some further research on these forums I have noted that other people have done the following:

* on the Media Server create a NET_BUFFER_SZ file in the following directory: install_path\Netbackup\db\config\
* add 132096 to the file (equivalent to 129K)
* set the client communication buffer size to same size as NET_BUFFER_SZ on the Media Server

Any ideas who's right?

Colin.
3 REPLIES 3

Chia_Tan_Beng
Level 6
Both are not wrong depending the whole picture of your environment.

If all your clients have the bottleneck (e.g. busy or slow disk) and couldn't improve in moving the data across the network, having a huge NET_BUFFER_SZ would give a huge number of waits.

But if your client can improve its performance by increasing its communication buffer size (i.e. configured under clients properties), increase it in conjunction with NET_BUFFER_SZ size would help. For relatively new servers, its normally the case.

Hope it helps.

Derek_Black_2
Level 3
Chia is right ... it depends on your environment. And both answers are right as well.

I usually see what the MTU is for the NIC and set the buffers to the same. This aligns the transmissions to coincide with the packet boundaries anyway and makes it more efficient. If your NIC/switch setttings specify a smaller MTU, then the NetBackup buffers will wait until they're full. If the MTUs are much bigger, then the NetBackup buffers are the bottleneck and need to be increased.

Hope this helps.

Stumpr2
Level 6
Source:
DOCUMENTATION: How to configure buffers for NetBackup in a Windows NT/2000 environment to improve performance
http://support.veritas.com/docs/244652


NET_BUFFER_SZ
Note the NET_BUFFER_SZ value is the size of the buffer on the media server which receives data from the client server. The NET_BUFFER_SZ is set on the media server by creating the NET_BUFFER_SZ file in the \NetBackup directory. The value in the file needs to be a multiple of 1024. If the file is not present, the default value of 262144 (or 256K) is being used. If the file is created for tuning purposes, be sure the file has no extensions.

NOTE: It is frequently helpful to match the Buffer_Size on the clients to NET_BUFFER_SZ on the media server.

Buffer_size
Please note the value of the Buffer_size does not show up in the bptm log file, but can be viewed and set from the master server or by modifying a registry value. The default value for the Buffer_size setting is 32K.

On the client server, the buffer size (Buffer_size) can be modified in two ways.

1. From the NetBackup Administration console, Host Properties | Clients, double-click the client you are configuring. Expand the options under Windows Client | select Client Settings. Change the Communications buffer to the desired size in KB.
2. Edit the following registry key:
HKEY_LOCAL_MACHINE\SOFTWARE\VERITAS\NetBackup\CurrentVersion\Config\Buffer_Size

Change the decimal value to the desired setting.

NOTE: Performance can sometimes be enhanced by setting this value the same as the NET_BUFFER_SZ on the media server.