Solution to the timeout scenario of Flink streaming computing (official live broadcast)

give an example ： Two streams , An order flow A collection stream
You can't put calculations in storage shortcoming ：

• The amount of detailed data is huge , Storage overhead
• The workload of product interface development is large
• Great promotion period , Data peaks , The double pressure of reading and writing
• OLAP The cost of the database itself is very high

Based on message queuing ：

• Delayed read （ Data timeout future manual manufacturing messages trigger ）
• Delayed distribution （ Data timeout future manual manufacturing messages trigger ）

be based on Flink State

• TimerService（ Data timeout future manual manufacturing messages trigger – It is best to ）
• CEP（ Data timeout future fetching and stream measurement processing ）