doi:10.1016/j.peva.2004.09.002
Copyright © 2004 Elsevier B.V. All rights reserved.
Analysis on packet resequencing for reliable network protocols
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Ye Xiaa, 
, 
and David Tseb
aComputer and Information Science and Engineering Department, University of Florida, Room 301, CSE Building, P.O. Box 116120, Gainesville, FL 32611-6120, USA
bElectrical Engineering and Computer Science Department, University of California, 261M Cory Hall, Berkeley, CA 94720-1770, USA
Received 3 April 2004;
revised 22 September 2004.
Available online 11 November 2004.
Abstract
Packets are sometimes disordered in the network. Reliable protocols such as TCP require packets to be accepted, i.e., delivered to the receiving application, in the order they are transmitted at the sender. In order to do so, the receiver’s transport layer must resequence the packets with the help of a resequencing buffer. Even if the application can consume the packets infinitely fast, the packets may still be delayed for resequencing. In this paper, we model packet disordering by adding an independently and identically distributed (IID) random propagation delay to each packet and analyze the required buffer size for packet resequencing and the resequencing delay for an average packet. We demonstrate that these two quantities can be significant and show how they scale with the network bandwidth.
Keywords: Resequencing queue; Packet disordering; Transmission control protocol
Fig. 1. The network model.
Fig. 2. Resequencing network model followed by a GI/GI/1 queue.
Fig. 3. Acceptance ratio vs. buffer size for block sizes: (a) n=10 ; (b) n=100.
Fig. 4. Acceptance ratio converges to 50%: buffer size b=0.5n.
Fig. 5. Selective-repeat ARQ examples. Two cases for the buffer size: b=1 (left) and b=2 (right).
Fig. 6. A snapshot of the resequecing buffer.
Fig. 7. Tail probabilities for 
(“numerical” and “approximation”) and loss probabilities under finite buffer size (“simulation”), for the exponential distribution (1/λ=20 ms): (a) Cs=1 Mbps; (b) Cs=100 Mbps.
Fig. 8. Tail probabilities for 
for the exponential distribution. Comparison of different mean delays. Cs=100 Mbps.
Fig. 9. Tail probabilities for 
for the Pareto distribution (α=1.9): (a) Cs=10 Mbps; (b) Cs=100 Mbps.
Fig. 10. Tail probabilities for 
for the Pareto distribution (“numerical” and “approximation”): comparison with actual loss probabilities under finite buffer sizes (“simulation”). Cs=1 Mbps; α=1.9 ; d=1000 s.
Fig. 11. Tail probabilities for 
for the Pareto distribution (log–log plot): (a) α=1.9, mean delay = 20 ms; (b) Cs=1 Mbps, mean delay = 20 ms; (c) Cs=100 Mbps, α=1.9, mean delays = 10 and 20 ms.
Fig. 13. Throughput vs. buffer size for “truncated” Pareto distributions (α=1.9, T=30 ms): (a) Cs=10 Mbps, linear scale; (b) Cs=10 Mbps, semi-log scale; (c) Cs=100 Mbps, linear scale; (d) Cs=100 Mbps, semi-log scale.
Fig. 14. Throughput vs. buffer size for “truncated” Pareto distributions (α=1.1, T=30 ms): (a) Cs=1 Mbps, linear scale; (b) Cs=1 Mbps, semi-log scale; (c) Cs=10 Mbps, linear scale; (d) Cs=10 Mbps, semi-log scale.
Table 3.
Lower bound on k*−k to achieve P{M(t)=k}≤
for the Pareto distribution
Mean delay = 20 ms, α=1.1.
Table 4.
Mean waiting time for the exponential delay: analysis vs. simulation
Table 5.
Mean waiting time for Pareto delay: simulation results
Corresponding author. Tel.: +1 352 392 2714; fax: +1 352 392 1220.
Abstract
Purchase PDF (1157 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (3)
Purchase PDF (5492 K)
