I have a requirement that will use Flink to parse data and put it into a DB (like Doris, CH). The program will do a simple statistical analysis of the data before entering it into the DB. For example, counting the number of occurrences of a certain IP, and always adding up. And write the statistics to the business DB (like PG, Mysql, etc...) periodically. (The results of statistical analysis do not have particularly high real-time requirements).
I understand that the above operation can be done by accumulator, metrics system, and aggregator function. Which one would be the optimal option in terms of performance? Also, when using an accumulator in CONTINUOUS_UNBOUNDED Flow, how do I periodically get the results of the accumulator?
