on 06-10-2014 10:03 AM
Hello,
i intend to use a HANA instance as HANA-based accelerator to accelerate some time-costly db reading transactions, using a secondary db connect.
I'm examining the SAP LT Replication Server for real-time data replication to HANA, for having my whole process of read and update data untouched, but transform the costly selection statements to use a secondary db connection.
While playing around, i noticed that the replication process isn't handled in real-time, and subsequent reading transactions operate on the non-replicated data set.
In my use case, this is a crucial problem.
Is there an appliance to monitor the replication process, so I can detect the finishing of the process? Is there a signal that flags the finishing of the replication process?
Have you any other approach to disrupt and continue my transaction sequence, to assure to operate on the most actual data?
Thanks, Stephan
Maybe my expactations where too strict, since near-real-time is not too bad.
However for my business case, finishing the replacation process before triggering the next sql request on the secondary db connection was crucical, so I implemented a logic, that checked the shadow tables, that none data is left for replication.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Pawel,
there are plenty of tables, yet for my scenario, detect if a data model, consisting of several DB tables, is replicated, I found the following tables useful and sufficient:
In the end, you just need to check that none of the shadow tables contains entries anymore, then your replication process can be considered as finished.
Hope this helps,
Stephan
Hi Stephan,
why is this not real time for you? What are your latency times?
Depending on your sizing and the change rate on the source system, SLT can achieve transfer rates below 1 second. The replication itself is an ongoing process, permanently looking if there is a change in the source.
Best,
Tobias
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Tobias,
thanks for your reply.
To me, real-time would imply, that, at the end of my db transaction, the applied data record is already replicated. Yet the LT replication server offers a real-time launched replication, but that's ok if i can deal with it in a smart way.
In my usage scenario, accelerate extreme transaction processing, i have two subsequent transactions, both operating on the same data record. Transaction one updates the data record on the primary application db, which is to be replicated to a HANA db. Transaction two selects the data record via secondary db connection from the HANA db and returns the pre-replicated data record. I have average latency times below 1 second on this scenario.
For me, it would be sufficient to pause my process until the replication process is finished, the overall acceleration due to the HANA select makes it worth, but i need some kind of signal to continue my process and perform the HANA select on the most recent data set.
Regards, Stephan
User | Count |
---|---|
93 | |
11 | |
10 | |
9 | |
9 | |
7 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.