All instances are SQL Server 2014 Ent. edition
I would appreciate your input and thoughts about this design.
We have set this design at our lab and so far done some very basic tests simulating fail overs etc.
However, due to the design’s complexity I am interested to hear other opinions.
Here is a link to see a diagarm of the design:
https://drive.google.com/file/d/0B7yeAe4Hpq4SZEVVc1JjM3VHOXc/view?usp=sharing
The diagram shows the following:
sql03 + sql04 are both FCI (Failover Cluster Instance) serving as active/passive nodes and holding the OLTP database (most crucial db).
sql05, sql06 and sql07 each hold 2 databases that receive data via Transactional Replication (these 2 databases on each instance are the subscription databases of a 1-way Transactional Replication)
The OLTP database (hosted by the FCI on sql03 and sql04) is the Primary Replica of AG-OLTP sending data to the Secondary Replicas hosted by sql01 + sql02
The subscription databases (hosted by sql05, sql06, sql07) are each the Primary Replica for its 2 databases hosted by AG-PTS01, AG-PTS06 and AG-PTS07. sending data to the Secondary Replicas hosted by sql01 + sql02
So sql01 + sql02 receive data from the various sources via the different AGs
These 2 servers are used for reports. They hold 2 partitioned tables that have the same schemes but one is a column store table and the other is a b-tree table (different type of queries will access each table, aggregates and none aggregates).
For load balancing and HA sql01 and sql02 each send data to the other via AG-REP01 and AG-REP02 respectively allowing applications to access both servers.
Thank you
-Yaniv
Yaniv Etrogi
site |
blog | linked in |
mail
Please click the Mark as Answer button if a post solves your problem! orVote As Helpful