cancel
Showing results forΒ 
Search instead forΒ 
Did you mean:Β 

Restrore speed uneven when restoring mssql

lafo
Level 4

Hi.

Our restorespeed is uneven. Looks lik the pitcure below. Master is Windows 2008. Media is disk (datadomain). Can anyone spread some light on this, why the restore isn't running at a staedy speed ?

Regards

Lasse

7 REPLIES 7

lafo
Level 4

Sorry, no pitcure.

The pitcure is like a wave, goes from zero to about 50 %, stays ther for 1-2 seconds, goes back to 0 stay for 12 seconds and then back to 50%.... and so on

 

//Lasse

Mark_Solutions
Level 6
Partner Accredited Certified

Not much information but sounds like a buffering issue

What version of NBU, do you have any NET_BUFFER_SZ, of NUMBER/SIZE_DATA_BUFFERS/_DISK in place?

When the job completes does it report how many times it was delayed etc.

bptm log from the Media Server may assist

lafo
Level 4

Hi

Netbackup ver 7.1

NUMBER_DATA_BUFFERS = 48

NUMBER_DATA_BUFFERS_DISK = 48

Info bptm(pid=6800) waited for empty buffer 4 times, delayed 228 times

bptm log is attached

Regards

//Lasse

Mark_Solutions
Level 6
Partner Accredited Certified

AH! Not a SQL backup, just a file system backup of SQL dumps

What SIZE_DATA_BUFFERS(_DISK) do you use?

Also, what fragment size do you use on the Storage Unit?

Yasuhisa_Ishika
Level 6
Partner Accredited Certified

Did you restored MS SQL Server database from backup taken with MS-SQL-Server policy?

In general, write is slower than read, especially while writing file data into RAID volume and writing data through applications and databases. It is usual.

Under some particular conditions, you may observe symmetric performance characteristics. But it is just because bottleneck point in data flow has such characteristics(e.g. Network). If bottleneck point has asymmetric performance characteristics(e.g. disk, application), write speed is not equal with read speed.

lafo
Level 4

HI.

Thanks for the replies.

It seems to be the same behavour even when we do a simple file-restore.

Performance is poor. Mabye our datadomain is overloaded.

I will investigate.

//Lasse

Mark_Solutions
Level 6
Partner Accredited Certified

Check your Buffer Sizes and reduce your Storage Unit fragment size to 5000MB - this often helps with restores from de-duplication data