none
DPM Concept - Synchronization RRS feed

  • Question

  • Hello all

    I'm struggling with a DPM concept I hope somebody can help me with.

    As I understand it, DPM makes a replica of the data you want to protect which is then stored on disc (or tape) as a recovery point. Changes to the protected data is sychronized to the replica as per the schedule.

    In my scenario, I have a volume in a Protection Group that makes three recovery points a day, at 08:00, 12:00 and 18:00 but makes synchronizations every 15 minutes. My understanding was that lets say at 08:00 my recovery point is made to disc. If I then make changes to the protected data, should those changes be applied to the recovery point every 15 minutes? E.g. if I change a text file and then wait for the next synchronization should I be able to recover the latest file? I can only recover the file as it was at 08:00 - no recoverable changes are made until the next recovery point at 12:00 is made to disc. Therefore what is the point of a 15-minute synchronization period if I can only recover from one of the three recovery points?

    Sorry if I haven't explained that well.

     

    Wednesday, October 19, 2011 12:57 PM

Answers

  • One purpose of doing multiple, more frequent synchronizations is to reduce the amount of time it takes to synchronize right before a recovery point.  If you only synchronized right before a recovery point, much more data would have to be transmitted over the wire at that point in time.  This can make a big difference especially over WAN connections.

    Another reason is it can enhance your restore options. The underlying replica is updated by the synchronization process.  You can recover outside of the normal recovery point schedule in at least two ways: create a one-off recovery point, or access the replica directly and copy the needed data out of it.

    Example scenario:  You synchronize your data every hour, but only take recovery points at 8am, 12pm, and 6pm.  Someone comes to you (the DPM admin) at 9:15 and says they just deleted a ton of data they created less than an hour ago.  You can't restore from the 8am recovery point because the data did not exist at that time. But it would have been synchronized to DPM at 9am.  You can go to the DPM GUI and create a one-off recovery point (without synchronization!) and then recover the data as it was at 9am (the last synchronization).

    If you only synchronize right before a normal recovery point, you wouldn't be able to do this.  This is one reason frequent synchronizations are generally a good idea.



    • Edited by Rod Savard Wednesday, October 19, 2011 1:51 PM
    • Marked as answer by Ian_Sun Thursday, October 20, 2011 6:57 AM
    Wednesday, October 19, 2011 1:50 PM

All replies

  • Hi Ian,

    the first inital full backup is the replica - this not a recovery point!
    A recovery point is created by an express full backup

    "(An express full backup is a type of synchronization in which the protection agent transfers a snapshot of all blocks that have changed since the previous express full backup (or since the initial replica creation, for the first express full backup) and updates the replica to include the changed blocks. The impact of an express full backup operation on performance and time is expected to be less than the impact of a full backup because DPM transfers only the blocks changed since the last express full backup.)"

    You are right, synchronization updates the replica (not the recovery point!)
    Normally you can only restore (file) data from recovery points BUT in case of DR you can manually create a recovery point from the last replica+syncs - this is usually done when the protected server is unavailable. -> Then your restored files are max 15 min old.
    Normal restore is made from recovery points.

    This is only for file data, appliacations (EXCH, SQL, Sharepoint) are handled differently.

    I hope this helps a little bit

    regards

    /bkpfast

     

    Wednesday, October 19, 2011 1:46 PM
  • One purpose of doing multiple, more frequent synchronizations is to reduce the amount of time it takes to synchronize right before a recovery point.  If you only synchronized right before a recovery point, much more data would have to be transmitted over the wire at that point in time.  This can make a big difference especially over WAN connections.

    Another reason is it can enhance your restore options. The underlying replica is updated by the synchronization process.  You can recover outside of the normal recovery point schedule in at least two ways: create a one-off recovery point, or access the replica directly and copy the needed data out of it.

    Example scenario:  You synchronize your data every hour, but only take recovery points at 8am, 12pm, and 6pm.  Someone comes to you (the DPM admin) at 9:15 and says they just deleted a ton of data they created less than an hour ago.  You can't restore from the 8am recovery point because the data did not exist at that time. But it would have been synchronized to DPM at 9am.  You can go to the DPM GUI and create a one-off recovery point (without synchronization!) and then recover the data as it was at 9am (the last synchronization).

    If you only synchronize right before a normal recovery point, you wouldn't be able to do this.  This is one reason frequent synchronizations are generally a good idea.



    • Edited by Rod Savard Wednesday, October 19, 2011 1:51 PM
    • Marked as answer by Ian_Sun Thursday, October 20, 2011 6:57 AM
    Wednesday, October 19, 2011 1:50 PM
  • Hmm, kind of. I'm just a bit hazy on why I would synchronize every 15 minutes if I can't restore to the same granularity. It would seem to only make sense to synchronize just before a recovery point which is one of the options.

    Ian

     

    Wednesday, October 19, 2011 1:50 PM
  • Because if you have alot of data then you will overload your network by copying all the changes at once. While if you sync more often then data will be moved from the server into DPM database (block wise) and then when the time comes to create recovery point DPM combine all sync data into one recovery point.

     

    Hope that helps,

    Laith.

    __________________________________________________-

    If my answer helped you please mark it as an Answer

    Wednesday, October 19, 2011 8:04 PM
  • One purpose of doing multiple, more frequent synchronizations is to reduce the amount of time it takes to synchronize right before a recovery point.  If you only synchronized right before a recovery point, much more data would have to be transmitted over the wire at that point in time.  This can make a big difference especially over WAN connections.

    Another reason is it can enhance your restore options. The underlying replica is updated by the synchronization process.  You can recover outside of the normal recovery point schedule in at least two ways: create a one-off recovery point, or access the replica directly and copy the needed data out of it.

    Example scenario:  You synchronize your data every hour, but only take recovery points at 8am, 12pm, and 6pm.  Someone comes to you (the DPM admin) at 9:15 and says they just deleted a ton of data they created less than an hour ago.  You can't restore from the 8am recovery point because the data did not exist at that time. But it would have been synchronized to DPM at 9am.  You can go to the DPM GUI and create a one-off recovery point (without synchronization!) and then recover the data as it was at 9am (the last synchronization).

    If you only synchronize right before a normal recovery point, you wouldn't be able to do this.  This is one reason frequent synchronizations are generally a good idea.




    Thanks to everyone - Rod's answer explains it perfectly. I hadn't realised that creating a one-off recovery point would allow me to do that. Thanks very much. :)

     

    Thursday, October 20, 2011 6:58 AM
  • Example scenario:  You synchronize your data every hour, but only take recovery points at 8am, 12pm, and 6pm.  Someone comes to you (the DPM admin) at 9:15 and says they just deleted a ton of data they created less than an hour ago.  You can't restore from the 8am recovery point because the data did not exist at that time. But it would have been synchronized to DPM at 9am.  You can go to the DPM GUI and create a one-off recovery point (without synchronization!) and then recover the data as it was at 9am (the last synchronization).

     

    Does that follow then that if the user came to me at 10:15 then they WOULD have lost the data? I.E. the recovery point is made at 8:00am, a sync is done at 9:00am and then the user deletes their data at 09:15am but they don't tell me. At 10:00am another scheduled sync is done so the data is now permanently lost as the sync processed sync'd the lost data?

     

    Friday, October 21, 2011 8:06 AM