Estimating the packet loss ratio of TCP transfers is essential for passively measuring Quality of Service (QoS) on the Internet traffic. However, only a few studies have been conducted on this issue. The Benko-Veres algorithm is one technique for estimating the packet loss ratio of two networks separated by a measurement point. However, this study shows that it leads to an estimation error of a few hundred percent in the particular environment where the packet loss probabilities between the two networks are asymmetrical. We propose a passive method for packet loss estimation that offers improved estimation accuracy by introducing classification conditions for the TCP retransmission timeout. An experiment shows that our proposed algorithm suppressed the maximum estimation error to less than 15%.
Takuya TOJO
NTT Corporation
Hiroyuki KITADA
NTT Corporation
Kimihide MATSUMOTO
NTT Corporation
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Takuya TOJO, Hiroyuki KITADA, Kimihide MATSUMOTO, "Passive Method for Estimating Packet Loss Ratio Based on TCP Retransmission Behavior" in IEICE TRANSACTIONS on Communications,
vol. E96-B, no. 7, pp. 1908-1917, July 2013, doi: 10.1587/transcom.E96.B.1908.
Abstract: Estimating the packet loss ratio of TCP transfers is essential for passively measuring Quality of Service (QoS) on the Internet traffic. However, only a few studies have been conducted on this issue. The Benko-Veres algorithm is one technique for estimating the packet loss ratio of two networks separated by a measurement point. However, this study shows that it leads to an estimation error of a few hundred percent in the particular environment where the packet loss probabilities between the two networks are asymmetrical. We propose a passive method for packet loss estimation that offers improved estimation accuracy by introducing classification conditions for the TCP retransmission timeout. An experiment shows that our proposed algorithm suppressed the maximum estimation error to less than 15%.
URL: https://global.ieice.org/en_transactions/communications/10.1587/transcom.E96.B.1908/_p
Copy
@ARTICLE{e96-b_7_1908,
author={Takuya TOJO, Hiroyuki KITADA, Kimihide MATSUMOTO, },
journal={IEICE TRANSACTIONS on Communications},
title={Passive Method for Estimating Packet Loss Ratio Based on TCP Retransmission Behavior},
year={2013},
volume={E96-B},
number={7},
pages={1908-1917},
abstract={Estimating the packet loss ratio of TCP transfers is essential for passively measuring Quality of Service (QoS) on the Internet traffic. However, only a few studies have been conducted on this issue. The Benko-Veres algorithm is one technique for estimating the packet loss ratio of two networks separated by a measurement point. However, this study shows that it leads to an estimation error of a few hundred percent in the particular environment where the packet loss probabilities between the two networks are asymmetrical. We propose a passive method for packet loss estimation that offers improved estimation accuracy by introducing classification conditions for the TCP retransmission timeout. An experiment shows that our proposed algorithm suppressed the maximum estimation error to less than 15%.},
keywords={},
doi={10.1587/transcom.E96.B.1908},
ISSN={1745-1345},
month={July},}
Copy
TY - JOUR
TI - Passive Method for Estimating Packet Loss Ratio Based on TCP Retransmission Behavior
T2 - IEICE TRANSACTIONS on Communications
SP - 1908
EP - 1917
AU - Takuya TOJO
AU - Hiroyuki KITADA
AU - Kimihide MATSUMOTO
PY - 2013
DO - 10.1587/transcom.E96.B.1908
JO - IEICE TRANSACTIONS on Communications
SN - 1745-1345
VL - E96-B
IS - 7
JA - IEICE TRANSACTIONS on Communications
Y1 - July 2013
AB - Estimating the packet loss ratio of TCP transfers is essential for passively measuring Quality of Service (QoS) on the Internet traffic. However, only a few studies have been conducted on this issue. The Benko-Veres algorithm is one technique for estimating the packet loss ratio of two networks separated by a measurement point. However, this study shows that it leads to an estimation error of a few hundred percent in the particular environment where the packet loss probabilities between the two networks are asymmetrical. We propose a passive method for packet loss estimation that offers improved estimation accuracy by introducing classification conditions for the TCP retransmission timeout. An experiment shows that our proposed algorithm suppressed the maximum estimation error to less than 15%.
ER -