1) What are t0/t1/t2/t3

IEC104 has its own link confirmation and keep-alive mechanism on top of TCP, typical parameters are:

• t0: Connection/Confirmation timeout (Establishment/Handshake phase)
• t1: I-frame confirmation timeout (Wait for confirmation after sending I-frame)
• t2: S-frame confirmation aggregation delay (Allow short-time aggregation confirmation, reduce ack frequency)
• t3: Idle test frame period (Send test frame when link is idle for a long time)

Tuning Suggestions:

• High network jitter on site: Appropriately increase t1/t0
• Need to quickly detect disconnection: Appropriately decrease t3 (But will increase heartbeat overhead)

2) k/w Window and Throughput

• kWindow: Max number of "Unconfirmed" I-frames allowed (Sliding window)
• wThreshold: Trigger confirmation after receiving how many I-frames (Or trigger confirmation aggregation logic)

General Rule:

• Larger kWindow allows more in-flight frames, higher throughput limit
• But if the window is too large and the peer cannot handle it, backlog, timeout and retransmission risks will occur

Suggest starting with default values and observing:

• Average response time, failure rate
• Whether there is discard/backpressure caused by "Window Full"

3) Backpressure Parameters: Protecting Gateway and Field Devices

The driver includes a set of "Memory Budget/Queue Capacity" parameters to protect the system under abnormal conditions:

• sendQueueCapacity: Send queue capacity (Avoid infinite accumulation)
• maxPendingAsduBytes: Pending ASDU bytes limit (Prevent memory from being blown up by uplink flood)
• discardLowPriorityWhenWindowFull: Discard low priority when window is full (Protect critical data)
• mergeLowPriority: Merge low priority (Keep only the last value, reduce queue pressure)
• lowPrioFlushMaxAgeMs: Allow low priority delayed flush (Reduce sending frequency caused by jitter)

Best Practice

• "Critical Telesignaling/Telecontrol Feedback" priority should be high (Decided by protocol side), ensuring critical data is preserved as much as possible under backpressure strategy.
• For high-frequency changing telemetry, recommend letting core use report_type=Change for change filtering, and avoid filling up the forwarding queue through merging/throttling on the southward side.