Receive Agent Node Buffer Management for NoC Congestion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network congestion and deadlock occur in computer networks and networks-on-chip due to data traffic exceeding the processing capacity of network devices, leading to degraded performance, increased latency, and potential data loss.
Innovation Solution
A data processing method that determines the number of available buffer units in a request receiving buffer and discards request data packets when the available buffer units are less than or equal to a threshold, implementing an internal retry mechanism to ensure available buffer units for subsequent data packets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data traffic is increased to meet higher throughput requirements, then network throughput is improved, but network congestion or deadlock occurs
Solution Approach 1:
The system performs preliminary buffer availability checks before accepting data packets. The receive agent node determines the number of available buffer units in advance, and only accepts packets when sufficient buffer space exists, preventing congestion before it occurs
Solution Approach 2:
The system implements a feedback mechanism where the receive agent node monitors buffer availability and sends credit information back to the request agent node. This feedback loop allows the system to dynamically adjust packet acceptance based on current buffer status, preventing both congestion and unnecessary retransmissions
2Reliability
If buffer units are reserved to prevent congestion, then network reliability is improved, but data transmission delay increases
Solution Approach 1:
The system dynamically adjusts buffer allocation and packet acceptance based on real-time buffer availability. Rather than statically reserving buffers, the receive agent node continuously monitors and reports available buffer units, allowing the network to adaptively utilize available capacity while preventing congestion
Solution Approach 2:
The system changes the parameter of buffer availability threshold dynamically. By adjusting the first threshold based on current network conditions and buffer status, the system can optimize between preventing congestion and minimizing transmission delay under different operating conditions
3Reliability
If request data packets are discarded when buffer units are insufficient, then network congestion is prevented, but data loss occurs
Solution Approach 1:
The system implements a discard-and-recover mechanism where packets are temporarily discarded when buffers are full, but the receive agent node tracks these discarded packets and requests retransmission when buffer space becomes available. This ensures no permanent data loss while preventing congestion
Solution Approach 2:
The feedback mechanism includes credit-based flow control where the receive agent node informs the request agent node about buffer availability. When buffers are full, transmission is paused; when space becomes available, credits are sent back to resume transmission, ensuring packets are only discarded temporarily rather than permanently lost
Data Source
AI summary
Embodiments of the present disclosure provide a data processing method, an electronic device, and a storage medium. The data processing method includes: determining, in a receive agent node in response to receiving a first request data packet, a number of available buffer units in a request receiving buffer used to receive the first request data packet; and discarding, in response to the number of available buffer units being less than or equal to a first threshold, at least one request data packet stored in the request receiving buffer. The method can ensure that there is at least one available buffer unit in a buffer of the receive agent node for receiving or storing requests, thereby preventing congestion or deadlock in an interconnect network-on-chip and improving the performance of the interconnect network-on-chip.