Analytical modeling of fast retransmission and recovery in TCP-RENO
Tsang‐Ling Sheu, Lien‐Wen Wu
Abstract
Tsang‐Ling Sheu, Lien‐Wen Wu
Abstract
This paper presents an analytical model to analyze the characteristics of fast retransmission and recovery when packets are lost in a TCP/IP network. In the model, we consider multiple packets that may be lost in a TCP congestion window, which makes this analysis technically different from the previous models as described in N. Caldwell et al. (2000), J. Padhye et al. (2000), B. Sikdar et al. (2003) and B. Sikdar et al. (2001). One of the significant observation revealed by this paper is that whether or not the lost packets can be fast retransmitted at the (y+1)-th RTT (round-trip time) round simply depends on N,[y], representing the number of packets followed by the first lost packet and were successfully received at receiver at th y-th RTT round. For clarification, the analytical model consists of two cases. The first case assumes that N,[y] is greater than or equal to a threshold k (the number of duplicate acknowledgments), while the second case assumes that N,[y] is less than the threshold k. From the analyses, we have found out that there are certain necessary conditions for the occurrence of fast retransmission, and we have derived an upper bound of the number of lost packets to be successfully recovered using the fast retransmission algorithm.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper presents an analytical model to analyze the characteristics of fast retransmission and recovery when packets are lost in a TCP/IP network. In the model, we consider multiple packets that may be lost in a TCP congestion window, which makes this analysis technically different from the previous models as described in N. Caldwell et al. (2000), J. Padhye et al. (2000), B. Sikdar et al. (2003) and B. Sikdar et al. (2001). One of the significant observation revealed by this paper is that whether or not the lost packets can be fast retransmitted at the (y+1)-th RTT (round-trip time) round simply depends on N,[y], representing the number of packets followed by the first lost packet and were successfully received at receiver at th y-th RTT round. For clarification, the analytical model consists of two cases. The first case assumes that N,[y] is greater than or equal to a threshold k (the number of duplicate acknowledgments), while the second case assumes that N,[y] is less than the threshold k. From the analyses, we have found out that there are certain necessary conditions for the occurrence of fast retransmission, and we have derived an upper bound of the number of lost packets to be successfully recovered using the fast retransmission algorithm.
Key concepts: Retransmission, Network packet, Computer science, Computer network, Upper and lower bounds, Algorithm, Real-time computing, Mathematics