Forum Discussion

JollyJ's avatar
JollyJ
Level 4
11 years ago

Restore is running with very Low throughput.

Master server: Window 2008,       NBU :7.5.0.4

Media server : Solaris 10,  NBU :7.5.0.4

DR

Master server: Window 2008,       NBU :7.5.0.4

Client : Windows 2003(VM machine), NBU : 7.5.0.4

We are trying to windows client  on over DR site. Restore is running with very low throughput

Nearly imposible to restore huge data of 10 TB.

Data is backed on tape with Solaris Media server with SIZE_DATA_BUFFER of 1048576.  We did

Faced issue while importing tapes at DR site due to 1 MB block size, However we overcome it

By changing the block size on HBA card.

       Now we are facing challenges  with Restore. We did replaced NIC with 10  GB. But no use

In detailed Tab we do see buffer delay time.

6 Replies

  • Never seen anyone use a buffe size that large! normally 256 is opmimum

    Have you set the number of data buffers too?

  • Number of Buffer are 128,

    SIZE_DATA_BUFFERS  are set 1048576 under the guidence of Symantec techsupport, to bump up the speed

    on data domain, But it has adversly affected our DR site recovery.

     

    message in detailed Tab shows

     " waited for empty buffer 6042 times, delayed 16176 times"

     I thing there is no point it tunnig buffer parameters, If yes please let me know..

     

  • are you possibly restoring over the wan? Are you selecting proper copy number?

  • It is Over the LAN and these are imported image (Primary copy)

  • JollyJ what is your NUMBER_DATA_BUFFERS on the primary site ?

    Seems to me that more buffers might help, also think there was a special NUMBER file for restores

    Then there is the Buffer size on the Windows client, which defaults to 32KB sometimes 64, 128 or even 256 gives better throughput

    Regards

    Michael