Use Case

As a Node.js driver engineer
I want to take a closer look at socket write performance regressions
So that I can improve the performance of writing

User Impact

• All users, socket writes are on every driver path

Dependencies

• None

Unknowns

• What is causing writes to be slower after implementing the new socket layer?
  • Should we write in chunks? Currently, we construct whole messages
  • Is there a difference between resolving the write promise in the callback to write or using the return value of write to determine if the promise can already be resolved?

Acceptance Criteria

Implementation Requirements

• Try to determine what change caused the write regression
• Try implementing the suggestions in the unknowns to see if they can improve performance

Testing Requirements

• Use the performance build variant to confirm changes improved performance

Documentation Requirements

• None

Follow Up Requirements

• None