TCP RTT Measurement via Variable Packet Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data communication networks, the inability to discern between acknowledgments for originally transmitted and retransmitted packets leads to retransmission ambiguity, making it difficult to accurately measure round-trip time and adjust transmission parameters, especially in networks with varying conditions.
Innovation Solution
The method involves transmitting packets with different sizes during initial and retransmission, allowing the sender to differentiate between acknowledgments using unique byte sequence numbers, enabling accurate round-trip time calculation and avoiding retransmission avalanches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the same packet size is used for initial transmission and retransmission, then packet transmission is simple, but the sender cannot discern which transmission is being acknowledged leading to retransmission ambiguity
Solution Approach 1:
The patent applies local quality by making the packet size different for initial transmission versus retransmission. Specifically, initial packets use a first size while retransmitted packets use a second, smaller size. This local differentiation allows the receiver to distinguish between acknowledgment of initial transmission versus retransmission, resolving the ambiguity without requiring complex additional markers.
2Loss of information
If timestamps are added to packets to avoid retransmission ambiguity, then acknowledgment identification is improved, but bandwidth overhead increases significantly
Solution Approach 1:
The patent changes the packet size parameter to differentiate between initial transmissions and retransmissions. By using a smaller packet size for retransmissions, the system creates a distinguishable pattern that allows the receiver to identify the type of transmission without adding extra fields like timestamps. This parameter-based differentiation achieves the same information differentiation with minimal overhead.
3Reliability
If retransmission timer is used to detect lost packets, then packet loss detection is achieved, but the sender cannot distinguish between lost packets and retransmitted packets leading to incorrect RTT measurement
Solution Approach 1:
The patent makes the packet size dependent on whether it is an initial transmission or a retransmission. This local quality differentiation allows the sender to accurately measure RTT for both initial packets and retransmitted packets by observing the acknowledgment patterns. The smaller size of retransmitted packets creates a distinct acknowledgment signature that enables precise RTT measurement without confusion between lost and retransmitted packets.
Data Source
AI summary
Round-trip time estimates may be calculated from an acknowledgement of an original packet transmission or from an acknowledgment of a retransmission without ambiguity by using different segment sizes for the same packet during transmission and retransmission. By using different segment sizes, the retransmitted segment will be acknowledged with a different acknowledgement number. In one embodiment, a TCP sender uses segment sizes that are slightly smaller than the maximum segment size for the initial transmission (e.g. MSS-4 bytes), and uses maximally sized segments during retransmission. When one of the two transmissions is explicitly acknowledged, the TCP sender can then use the different byte sequence acknowledgement numbers to discern which transmission is being acknowledged and correctly measure the round-trip time. This also allows the TCP sender to initiate the NewReno fast recovery algorithm only upon receipt of a retransmission acknowledgment. This is especially useful, for example, during periods of rapidly increasing delays caused by deteriorating network conditions or increased congestion.