Hello,
I have a couple of questions: the first one is how to improve replication latency and the second one is how to measure replication lag (average, instant) correctly
1) Question regarding improving replication lag.
I read that it is possible to improve performance (decrease lag) by adding new extracts and replicats to process DB updates in parallel.
So the question how it is better to organize it?
As I understand I need use one extract that reads changes from redo logs and a few replicats that reads changes from trail files and use different ranges.
So the question is it possible to use ranges for extracts. Please give an example.
Also it would be good if you provide any other methods to improve performance. I use integrated capture with downstream deployment mode.
2) How is it better to measure replication lag?
As I understand ggsci shows instant/latest lag, but I think it is better to measure average values probably for example to compare different environments.
In my case I implemented a script that do 10000 inserts, 10000 updates and 10000 deletes into specific table and via using triggers I calculated average lag. For example average Insert takes 3 seconds, average Update takes 10 seconds, average Delete takes 9.
But if I increase the number of inserts/updates/deletes and it means that load of DB is more than it was before and average lag is increased.
So how to measure replication lag correctly. In major cases instant (latest) replication lag is 10-15 seconds on my environmrnt. Is it possible to decrease it until 3-5 seconds for example?