Forum Discussion

thomas_crowe's avatar
13 years ago

Question about MSDP and Automated Image Replication (AIR)

I am working with Netbackup 7.1 and researching AIR with MSDP.  I have come across the statement "If you are using MSDP for A.I.R, only the deduplicated data gets replicated. If the target already had the segments, it is not send again.".  I understand what the poster is saying at face value, however my question is how.

My concern is what bits get transmitted over the wire.  How does the destination MSDP know that the segments are unique or already stored in its pool until it has recieved and analyzed those segments?  Do you configure the source media server as a load-balancer for the destination MSDP?

 

Thanks!

  • Hi,

     

    Its just doing optimized duplication as you would within a NetBackup domain when using SLP's to duplicate from one MSDP to another MSDP (or PureDisk to PureDisk). Remember these are storage devices that work at a layer "below" netbackup. The storage devices talk to each other to find out, what's new, what's already stored, and what to send.

     

    You don't configure any load balancers. The setup should reflect a regular push configuration where only the local media server is listed/configured in each storage unit respectively.

2 Replies

  • Hi,

     

    Its just doing optimized duplication as you would within a NetBackup domain when using SLP's to duplicate from one MSDP to another MSDP (or PureDisk to PureDisk). Remember these are storage devices that work at a layer "below" netbackup. The storage devices talk to each other to find out, what's new, what's already stored, and what to send.

     

    You don't configure any load balancers. The setup should reflect a regular push configuration where only the local media server is listed/configured in each storage unit respectively.

  • The communication takes place at the storage server level.  In the case of MSDP the media server IS the storage server but for PDDO, OST and 50xx appliances the storage server is separate from the media server. 

    The behavior is the same for both inter-domain (optimized duplication) and intra-domain (auto image replication) communication.  (It's also very similar to client side deduplication during the initial backup.)  It all comes down to the signature associated with the data segment.  Each segment has a unique signature(kind of like the check sum if you like) that is stored on each storage device along with the segment.  (Exactly how the segment and signature are stored depends on the device.)

    During a duplication the source storage server sends each segment signature to the target storage server.  The target storage server checks to see if it already has that segment.  If it doesn't it asks the source storage server to send over the segment. 

    That's pretty much all there is to it.  No need for any fancy load balancing and the good thing is that because each signature is unique even where you have multiple sources duplicating to the same target you will never have to send the same segment twice.