Complexity: Hard / help wanted
[Problem]
Connection management and asynchronous network operations can create unnecessary task, allocation, and scheduling overhead at high connection counts.
[Proposed Change]
Investigate a lower-overhead non-blocking I/O architecture appropriate to Nimbus.
Potential approaches include SocketAsyncEventArgs, pooled I/O state, reusable buffers, or other event-driven designs.
[Implementation Note]
The proposed APIs and architecture are starting hypotheses. Profile Nimbus under representative connection loads before replacing the existing I/O model.
[Acceptance Criteria]
•Connection scalability improves under representative load.
•Allocation/task overhead is reduced.
•Connection ordering and failure handling remain correct.
•Resource usage remains bounded.
Complexity: Hard / help wanted
[Problem]
Connection management and asynchronous network operations can create unnecessary task, allocation, and scheduling overhead at high connection counts.
[Proposed Change]
Investigate a lower-overhead non-blocking I/O architecture appropriate to Nimbus.
Potential approaches include SocketAsyncEventArgs, pooled I/O state, reusable buffers, or other event-driven designs.
[Implementation Note]
The proposed APIs and architecture are starting hypotheses. Profile Nimbus under representative connection loads before replacing the existing I/O model.
[Acceptance Criteria]
•Connection scalability improves under representative load.
•Allocation/task overhead is reduced.
•Connection ordering and failure handling remain correct.
•Resource usage remains bounded.