This is the sort of thing that shooting from the hip would only coincidentally hit the optimal target.
On two little factors I will comment claiming my comments are probably general:
- On 7-8 inserts per transaction, the most important bit is that you make sure all the pieces local to the data base before the first insert takes place and that since your commit point is by batches the entire batch is local to the database before the first insert takes place. You do not want even a LAN hiccup (or gcc overload in the case of RAC) exposing you extension of your concurrency footprint and most especially any uniqueness constraints.
- Some sort of “Scaling to Infinity” dataflow should be employed with respect to any partitioning of the transaction database and information life cycle longer retention. Usually either initial insert time or transaction completion time is the best archiving boundary, but either way the initial insert hour, day, week, or month is usually the best partitioning strategy to facilitate swap empty partition or truncate partition as the pathway to your “archive” database (whether or not the data from these archived partitions is then loaded into, say, snowflake, or some other decision support and analytical support data respository).
All y’all should probably invest in specific outside resources to help you shape the data flow of your solution having read in on your specific and detailed requirements for at least: availability, disaster business continuation, critical queries your business needs answered, and the curve of rate of loss by time unavailable of each of the services these database services supply.
Good luck,
mwf
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of yudhi s
Sent: Saturday, December 09, 2023 5:26 AM
To: Oracle L
Subject: Design related question
Hello Listers,
I understand colleagues here having decades of experience in the database world, We want some guidance, if the below design looks okay for our customer use case.
We currently have financial systems transaction data streams to Oracle exadata(X9) on-premise. This database supports processing of 400million transactions per day. And again a single transaction for us is a combination of 7-8 inserts into different transaction tables with Indexes , unique constraints etc defined on those and they commit in batches(~1000 batch size) in the database. And this system persists data for ~6 months. We do have all sorts of OLAP(daily/monthly batch reports running) applications run on the same database along with some user facing UI applications showing customer transactions. So it's basically currently serving a hybrid workload and is one stop solution for all use cases.
Many of the applications are moving from on premise to AWS cloud as part of modernization journey and AWS being chosen cloud partner also the product is expected to expand across more regions and this system is expected to serve increase in the transaction volume. And also we have a requirement to persist transaction data for ~10years to have those available for analytics/data science use cases.
So the team is thinking of splitting it into two parts
1)OLTP type use case in which we will persist/write the transaction data faster and show it to the UI related apps , in near real time/quickest possible time. and this database will store Max 60-90 days of transaction data. Not sure if we have an option of Oracle exadata equivalent on AWS, so team planning of using/experimenting with Aurora postgres. Please correct me, if there are any other options we should use otherwise?
2)Then move the data beyond ~90 days into another database or object storage S3 which will keep it there for ~10 years and will be queryable using the necessary API's. That is supposed to cater to Olap/analytics/data science use cases etc.
Is the above design is okay? and also in regards to the second point above i.e. persisting the historical data (that to be in queryable state), should we go for some database like snowflake or should just keep it on S3 as is and make those queryable through APIs. Please advice?
Regards
