Selective Acknowledgement RDMA Packet Retransmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
RDMA protocols face inefficiency in packet retransmission due to the need to resend all packets following a single lost packet, leading to increased network load and performance degradation, especially in networks with significant packet loss.
Innovation Solution
Implementing a method for selective acknowledgment and retransmission in RDMA data transfer transactions, where a negative acknowledgment (NAK) packet is sent to retransmit only the missing packets, allowing the network to efficiently resume data transfer without retransmitting successfully received packets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all packets following a lost packet are retransmitted to ensure reliable delivery, then reliability is improved, but network bandwidth consumption increases and productivity deteriorates
Solution Approach 1:
The patent segments the acknowledgment mechanism into selective acknowledgments for individual packets rather than bulk retransmission. The receiver tracks each packet's receipt status independently and sends targeted retransmission requests only for missing packets, dividing the retransmission task into discrete packet-level operations that avoid unnecessary data resending.
Solution Approach 2:
The patent applies local quality by treating each packet's delivery status independently. Instead of uniform retransmission of all subsequent packets, the system applies differentiated treatment: packets that were successfully received are not retransmitted, while only the specific missing packet is retransmitted. This localized approach optimizes bandwidth usage by matching retransmission actions precisely to actual delivery failures.
2Reliability
If all packets following a lost packet are retransmitted to ensure complete data delivery, then reliability is improved, but bandwidth consumption increases
Solution Approach 1:
The patent extracts the problematic bulk retransmission behavior and replaces it with selective packet-level retransmission. By taking out the unnecessary retransmission of already-received packets and keeping only the essential retransmission of missing packets, the system maintains delivery completeness while eliminating wasted bandwidth consumption on redundant data transmissions.
Solution Approach 2:
The patent discards the conventional approach of retransmitting all packets after a loss and recovers bandwidth efficiency through selective retransmission. The receiver identifies which specific packets are missing and requests only those be retransmitted, discarding the wasteful practice of resending entire packet sequences when only individual packets were lost.
3Reliability
If standard RDMA retransmission is used to handle packet loss, then reliability is maintained, but the time to recover from packet loss increases
Solution Approach 1:
The patent implements preliminary action by having the receiver proactively track and identify missing packets as they occur during transmission, rather than waiting for timeout-based retransmission triggers. The receiver maintains state information about expected packets and immediately generates retransmission requests for missing packets, enabling faster recovery without waiting for standard timeout periods to elapse.
Solution Approach 2:
The patent implements feedback mechanisms where the receiver continuously monitors packet arrival and provides immediate feedback about missing packets through targeted retransmission requests. This feedback loop enables the transmitter to quickly identify and retransmit only the specific missing packets, significantly reducing recovery time compared to blind retransmission of all subsequent packets.
Data Source
AI summary
A method for data transfer includes transmitting a sequence of data packets, including at least a first packet and a second packet transmitted subsequently to the first packet, from a first computer over a network to a second computer in a single remote direct memory access (RDMA) data transfer transaction. Upon receipt of the second packet at the second computer without previously having received the first packet, a negative acknowledgment (NAK) packet is sent from the second computer over the network to the first computer, indicating that the first packet was not received. In response to the NAK packet, the first packet is retransmitted from the first computer to the second computer without retransmitting the second packet.