Dear Oracle Experts, I am looking for help with a design decision on a DR database, a DW database and its disaster recover database.
At our company, we are debating the usage of shareplex vs dataguard for disaster recovery. Below is the scenario
The current data (5tb) is in mainframe and project is to move it to oracle, setup dw and dr databases. So we will have database A (high OLTP, 5TB initial size, expected to grow fast, 8tb, 12tb in few months, on a powerful V-block blade, 200g ram, oracle 12c rac etc, this database A is the source of all 3 other databases), A_DW (datawarehouse db for A, internal use and reside in same data center), A_DR (disaster recovery instance of A in a different datacenter) and A_DW_DR (DR database for A_DW in same datacenter as A_DR). Design was done in the past and I am reviewing/modifying the design as I feel relying on a replication tool for DR is going to have issues in future.
The replication from A -> A_DW is set to happen thru shareplex (a uni-directional replication, no lobs or complex data types, hence shareplex was chosen over goldengate), at the same time another shareplex session will replicate data from A -> A_DR. And there is dataguard setup to keep A_DW and A_DW_DR in sync. Also another requirement of using shareplex between A->A_DR is the ability of shareplexed A_DR database, be open for occasional read access to verify somethings on the fly (without connecting to A). Hence Architects in the past have chosen shareplex (over goldengate or activedataguard on cost basis).
My experience with goldengate (any replication tool) in the past was always troublesome, job abends, conflicts, lags with constant DBA escalations etc. Hence I proposing to use Dataguard between A->A_DR and leave shareplex as is, between A->A_DW and also allow users to have the occasional "read access requirement" to connect to A_DW (instead of keeping A_DR open) or go with active dataguard between A->A_DR altogether.
We can setup the db's and do POC but its hard to do a real simuation of the real time traffic and the previous design might work fine if the traffic is not high on A.
We have a DBA that has done shareplex before and he seems to be of opinion that if shareplex has lag, then dataguard most likely will also have a lag. But based on the design of how replication tools work, wouldnt the replication tool be slow as it reads from the redologs and instead dataguard transmits the redolog to destination and apply it ? Also how about the vendor support of shareplex vs datagaurd ? I just feel comfortable as dataguard is so inbuilt to oracle, 1 vendor - 1 support, fast swith over, fail over features, block corruption recovery with dataguard etc.
Please advise 1. Advantages and disadvantages of using shareplex vs dataguard for DR.
2. Should we consider goldengate (vs shareplex),
3. Shareplex performance in general and lag issues on high OLTP systems 4.
4. Would you anticipate issues with the original design especially in terms of performance on A, lag built between A and the remaining 3 databases.
Sorry for the lengthy question and thankyou for reading it thru. My original question was 3 or 4 times longer but I tried my best to cut it down but still be able to explain the design and questions about it.
Thankyou very much.