Where does such a high RPS come from?
Java
Tell me about your experience
Why did you not use ZGC or Shenandoah?
How was the problem of re-executing side effects solved?
How was it possible to double the RPS?
What happens to unprocessed messages in DLT afterwards?
Describe the architecture of a billing application — does it constantly calculate or only at request?
What are accruals in Gazpromneft? What kind of system is it?
How is idempotency organized in your system?
How has memory consumption changed? Did you have to configure anything?
How do clients get notified that there is a problem with calculations?
Input data is provided OperationId TransactionId ISIN RcvID SndId Quantity OperationId - int, operation ID TransactionId - int, transaction ID ISIN - String, instrument ID RcvID - int, recipient client ID, -1 if withdrawal SndId - int, sender client ID, -1 if deposit Quantity - double, instrument quantity OperDate - String, date in [phone] format Input lines, for example 1 1 RUB [phone] USD [phone] USD [phone] RUB [phone] Calculate the quantity of each instrument for each client on the latest date [phone] USD [phone] USD [phone] RUB 100
Does the order of message processing affect the result? For example, if the crediting failed in DLT, can it be rolled back?
How was reactive recalculation triggered during status transitions and related side effects?
Are there any adjustments to events? For example, 10 rubles were credited, but it should have been 9. What should be done?
If the business requirement is for the record to be exactly 9 rubles, and not 10, what should be done?
How is the description of transitions between statuses organized?
What is DLT?
How does a client see the situation when a deposit has failed and they want to write off?