Best one-way replication option going forward

  • I'm looking for some insight into implementing replication with a SQL 2005 Standard database based on our needs as well as the fact that one major replication type is deprecated in SQL 2012 and forward. To start off with, let me briefly describe what our replication needs are and what replication method I think is best for that... and why we don't want to use it. I'll also state that I have never dealt with SQL replication, so I'm learning as I go here.

    We have a single SQL database server that drives all of our production data for our company. We are wanting to setup a second read-only database server that we replicate to to help load balance for some specific applications that will be fairly intense with the database. This second SQL server is always online, on the same network and as mentioned will be read only--no sending data back to the primary server. Updates need to happen as data changes on the primary server.

    It seems fairly obvious to me that based on the above that we should use updatable subscriptions for transactional replication. However, this feature is deprecated in SQL 2012. Our IT manager wants to use something that's future proof, so she said out with this method. When we'll ever upgrade our SQL 2005 Standard installation, who knows. It could be many years. The only other real method that seems viable is merge replication. That isn't deprecated in SQL 2012. Peer-to-peer is what Microsoft seems to says is the replacement for updatable subscriptions for transactional replication, but we're not going to upgrade to enterprise for this feature.

    My question is: what are the downsides for using merge replication for a scenario like the above where clearly it seems updatable subscriptions for transactional replication is better. Can we effectively make Merge Replication work in a similar fashion to updatable subscriptions for transactional replication and avoid updates being pushed to the publisher? Are there any performance issues to consider? ...or other caveats for that matter?

    Thanks for any insights on the matter.

  • Since you're doing a full database replication instead of selective, I'd actually recommend using mirror/snapshot. It's basically easy-mode transactional replication. Check it out; it's definately future proof as it's part of the HA-DR methodology SQL wants to implement.

    A mighty quick walkthrough of how to do the setup:

    http://weblogs.sqlteam.com/tarad/archive/2007/02/13/60091.aspx


    - Craig Farrell

    Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.

    For better assistance in answering your questions[/url] | Forum Netiquette
    For index/tuning help, follow these directions.[/url] |Tally Tables[/url]

    Twitter: @AnyWayDBA

  • Interesting--I had not really looked at or considered that option. I'll have to dig around for some more extensive information and howtos on doing this. The link you gave seems to indicate the mirrored DB wouldn't be usable because of being set to norecovery, but it just looks like I have to find the right setup with snapshots to get what I need.

    Thanks--this gives me a new rabbit trail to go down.

    EDIT: It looks like this option is out. Snapshots aren't a feature of SQL 2005 Standard edition. We won't be dropping the money on Enterprise--that I know.

  • ESpigle (8/16/2012)


    Interesting--I had not really looked at or considered that option. I'll have to dig around for some more extensive information and howtos on doing this. The link you gave seems to indicate the mirrored DB wouldn't be usable because of being set to norecovery, but it just looks like I have to find the right setup with snapshots to get what I need.

    Thanks--this gives me a new rabbit trail to go down.

    EDIT: It looks like this option is out. Snapshots aren't a feature of SQL 2005 Standard edition. We won't be dropping the money on Enterprise--that I know.

    My pleasure and dead on target. The mirror isn't directly usable. What you do is snapshot it at certain time periods to 'upkeep' it. Snapshots are usually pretty quick. If you want to get particularly fancy you setup a full warehouse alongside the mirror/snapshot setup. Every x period your updater runs, creates a new snapshot, evaluates the delta, and includes it into the reporting warehouse without user interuption. 🙂

    ... and I just caught your edit. Well, that sucks. Um, hrm. Log shipping then?

    EDIT: Btw, I'd avoid merge replication like the plague. If you NEED merge, it's the only way to go, but it's such a PITA that I'd avoid it unless absolutely necessary.


    - Craig Farrell

    Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.

    For better assistance in answering your questions[/url] | Forum Netiquette
    For index/tuning help, follow these directions.[/url] |Tally Tables[/url]

    Twitter: @AnyWayDBA

  • Evil Kraig F (8/16/2012)

    EDIT: Btw, I'd avoid merge replication like the plague. If you NEED merge, it's the only way to go, but it's such a PITA that I'd avoid it unless absolutely necessary.

    We don't really need it, per se. It was just one of the options we saw that was not being done away with for future SQL versions. Already with just some initial testing I can see we're going to have many roadblocks just to make it work base on various things with the database structure and setup.

  • How up to date does the copy need to be kept? Snapshot replication isn't going anywhere anytime soon as far as I know, and should work in standard.

    ... which, at that point, you might just look into backup/restores to the other server. Sorry, scrap that. Not that it's a bad idea (testing your restores is always good), just probably too much downtime if you aren't looking for 'one day behind' upkeep.


    - Craig Farrell

    Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.

    For better assistance in answering your questions[/url] | Forum Netiquette
    For index/tuning help, follow these directions.[/url] |Tally Tables[/url]

    Twitter: @AnyWayDBA

  • ESpigle (8/16/2012)


    We are wanting to setup a second read-only database server that we replicate to to help load balance for some specific applications that will be fairly intense with the database. This second SQL server is always online, on the same network and as mentioned will be read only--no sending data back to the primary server. Updates need to happen as data changes on the primary server.

    Reading your initial requirements I'm not sure why you need to configure transactional replication with updateable subscriptions. You have said that your replicated database will be read-only, so surely no need for an updateable subscription?

  • I'm still trying to get a grasp on all the ways to get data from one server to be matched on another for application load balancing, so what you say may be entirely true. I'm at the learning about the various ways to accomplish this stage, so I've definitely not settled on one way while I'm trying to figure out how it all works. I'm still researching my options that are available for a SQL 2005 Standard setup.

  • I don't know the details of how it all works but we use SAN replication on a realtime basis to send transactions to our DR site. It's awfully fast and requires nothing of the SQL server.

    --Jeff Moden


    RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
    First step towards the paradigm shift of writing Set Based code:
    ________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

    Change is inevitable... Change for the better is not.


    Helpful Links:
    How to post code problems
    How to Post Performance Problems
    Create a Tally Function (fnTally)

  • Do you actually have the replicated DB online Jeff? I've used a couple of flavours of SAN replication but the database is not actually mounted at the other end

  • MissTippsInOz (8/17/2012)


    ESpigle (8/16/2012)


    We are wanting to setup a second read-only database server that we replicate to to help load balance for some specific applications that will be fairly intense with the database. This second SQL server is always online, on the same network and as mentioned will be read only--no sending data back to the primary server. Updates need to happen as data changes on the primary server.

    Reading your initial requirements I'm not sure why you need to configure transactional replication with updateable subscriptions. You have said that your replicated database will be read-only, so surely no need for an updateable subscription?

    I see why you asked this now and that I didn't have a full understanding of there being two types of transactional replication. I see now that obviously the version with updatable subscriptions is meant more for a two way type setup. Clearly with me just needing one-way replication and no updates from the subscribers, standard transactional replication is all I need. That said--is standard transactional replication something that isn't being deprecated with future SQL version? Is it only the updatable subscription type of transactional replication which is being deprecated (per THIS MS note)? I'm still digging to find the exact answer to this, but I'm being lead to believe that standard transactional replication isn't being removed necessarily.

  • I'm not aware that replication, generally, is being deprecated any time soon. While there may be other, better options, in SQL 2012 to achieve your requirements, if you need to implement a solution now I'd say its sufficiently future-proof.

Viewing 12 posts - 1 through 11 (of 11 total)

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