Distrubted Availability Groups -- Is this the right idea?

  • Evening All,

    We are looking at/reseaching implementing Availability Groups in SQL Server 2016.

    We have two data centers with the same hardware and storage options in each. Licensing is not a concern either. Neither is SAN Space.

    I am thinking of designing three synchronous and auto fail over replicas in each data center with with asynchronous between the data centers.

    However both data centers will be active hosting different databases at the same time. Is this a bad idea?

    To illustrate, literally:
    DataCenter #1 will host UK, FR, and AUS data (and will have an application pointed at it), there is a read only replica for report load and a HA node. The Availability groups will be asynchronous sent to the OTHER data center (DataCenter #2) for DR purpose.

    DataCenter #2 however will  be hosting the primary data for IRE, ITA and OTHR, again synchronous replicas locally for HA and asynchronous to DataCenter#1 for DR.

    There is plenty of capacity. So performance is not a cocncern and the line between the two is dedicated , latency is not usually a problem BUT there has been a couple of glitches last more than 5 minutes in the last 18 months. Hence, I am keen for connectivity between the two DC's to not impact the overall performance.

    What do you think? Is this a ridiculous idea?

    Cheers
    Alex

  • alex.sqldba - Wednesday, February 1, 2017 9:26 AM

    Evening All,

    We are looking at/reseaching implementing Availability Groups in SQL Server 2016.

    We have two data centers with the same hardware and storage options in each. Licensing is not a concern either. Neither is SAN Space.

    I am thinking of designing three synchronous and auto fail over replicas in each data center with with asynchronous between the data centers.

    However both data centers will be active hosting different databases at the same time. Is this a bad idea?

    To illustrate, literally:
    DataCenter #1 will host UK, FR, and AUS data (and will have an application pointed at it), there is a read only replica for report load and a HA node. The Availability groups will be asynchronous sent to the OTHER data center (DataCenter #2) for DR purpose.

    DataCenter #2 however will  be hosting the primary data for IRE, ITA and OTHR, again synchronous replicas locally for HA and asynchronous to DataCenter#1 for DR.

    There is plenty of capacity. So performance is not a cocncern and the line between the two is dedicated , latency is not usually a problem BUT there has been a couple of glitches last more than 5 minutes in the last 18 months. Hence, I am keen for connectivity between the two DC's to not impact the overall performance.

    What do you think? Is this a ridiculous idea?

    Cheers
    Alex

    alex.sqldba
    I don't think this is a bad/ridiculous idea at all.  I'm assuming that the active databases will reside near the application servers that will use them the most.  In an implementation that I performed, much like yours but on a smaller scale, latency was around applications servers accessing data over the wire.  Working with the Application/Server admins we setup a topology that was similar to our database topology.

    Thomas LiddleSQL Server AdministratorVideo Blog - YouTubeWeb Blog - www.thomasliddledba.comTwitter - @thomasliddledbaFacebook - @thomasliddledba

  • Hi Thomas,

    Yes, spot on -- whilst we haven't quite got a plan of action for the app servers (there is a potential in house rewrite of an existing vendor app on the horizon) that would certainly the intended set up. Quite how that is achieved either redundant VM's as App servers in each Data Centre or something else has not been decided or even broached yet.

    With your set up Thomas, did you have different domains in each data centre or was it a single domain?

    Alex

  • alex.sqldba - Tuesday, February 7, 2017 9:40 AM

    Hi Thomas,

    Yes, spot on -- whilst we haven't quite got a plan of action for the app servers (there is a potential in house rewrite of an existing vendor app on the horizon) that would certainly the intended set up. Quite how that is achieved either redundant VM's as App servers in each Data Centre or something else has not been decided or even broached yet.

    With your set up Thomas, did you have different domains in each data centre or was it a single domain?

    Alex

    Alex
    The one thing to consider is that if you decide to go VM with your implementation you have other options to choose from with VM technology that might be cheaper and not as complex to implement.  I would check with your sysadmins.  As for my implementation (SQL2K12) we had a single domain.  If you will have multi-domains AND your sysadmins deploy DC's on VM's, one thing that I would suggest is to create an AD domain just for SQL and then put trusts in place to your other domains (if needed).  Some sysadmins might not like that (multiple domains to support) but if you are constrained by multi-domains it might be something to consider.

    Also, think about security.  SQL Server 2016 can support gMSA which is exciting in AG's.

    Thomas LiddleSQL Server AdministratorVideo Blog - YouTubeWeb Blog - www.thomasliddledba.comTwitter - @thomasliddledbaFacebook - @thomasliddledba

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply