Log Shipping vs Replication vs Mirroring

Currently we are using SQL Server 2008 R2 and looking to go forward with Denali.

We will have 25,000 users connecting to an AX environment (SQL DB and Cube) and could be up to 10,000 concurrent users.

For our solution we have decided to log ship the AX database every 5-15mins and use this replicated db for all the cube and most of the reporting connections to ensure that the main server never gets overloaded or any issues come with processing the certain cubes every 5-15mins.

My question is what is the best approach to get this data across to the "replicated" server?
1. Log Shipping:
     a. Pros; easy to set up, no need to have pk on each table;
     b. Dis; User currently running reports connected to replicated db will be disconnected when the db gets log shipped every 5-15mins.
2. Replication:
      a. Pro: very fast, no downtime
      b. Dis: Complicated to set up and easy to fail.
3. Mirror:
      a. Pro: fast
      b. Dis: can't connect whilst being mirrored.

Can anyone confirm which you would feel is the best approach with the pros and cons. From my understanding log shipping would be best however the major issue is users getting disconnected whilst running reports?

Who is Participating?
That's a very complex topic, without a one-size-fits-all solution.  Many articles are available on the web, but here's one with a pretty complete bullet list of the Distinctions.

My 2-cents, based on the premise that you are mainly wanting to distribute workload rather than improve availability.  And with that many users doing analytical queries, I can see why!

Mirroring is almost exclusively designed to improve availability by providing a hot standby.  It is not intended for distributing workload and does not support more than one secondary server. However, if you are replicating to lots of secondaries, using a mirror as the primary for replication is a nice way to offload the replication overhead.

Log shipping was also designed as an availability solution.  Normally, it's used with the secondary offline until it is needed. Yes, it can also be configured to incrementally "clone" a database: a sort-of KISS one-way replication, one-way replication, so to speak. But if you are seriously considering synching your secondary every 5-15 minutes isn't the frequent disruption to thousands of users a show-stopper?

You mention "no need to have a PK on each table" as an advantage of log shipping. Not sure what you mean by that and, at face value, it does not sound at all like an advantage to me.  That said, I do agree that identity columns often used as primary keys present special challenges for replication processes.  However, since your secondaries would be read-only subscribers, the matter is easy to handle.

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.