TCP congestion control has a critical impact on the Internet stability and performance. In relatively recent times, a number of novel TCP congestion control variants, such as TCP BIC, CUBIC and Compound, started to be deployed in modern operating systems. While risks and benefits of these recent developments are subject of debate, the research community needs to monitor the extent to which each of these novel TCP congestion control variants is actually used in regulating Internet traffic, and this information, due to the complex, diverse, and decentralized nature of the Internet, is very difficult to be obtained. As a first step towards the collection of this important piece of information, we have developed a tool, presented in this paper, which allows us to identify the TCP variants employed in monitored TCP sessions. Our approach consists in passively monitoring packets and ACKs exchanged in TCP sessions, estimating the evolution of congestion window, and matching a set of features against those typical of the various TCP variants considered. Our identification tool has been validated trough emulation and real Internet experiments, showing promising results in the identification of New Reno, BIC, Cubic and Compound TCP variants.
TCPMoon: Monitoring the Diffusion of TCP Congestion Control Variants in the Internet
Granelli, Fabrizio;
2011-01-01
Abstract
TCP congestion control has a critical impact on the Internet stability and performance. In relatively recent times, a number of novel TCP congestion control variants, such as TCP BIC, CUBIC and Compound, started to be deployed in modern operating systems. While risks and benefits of these recent developments are subject of debate, the research community needs to monitor the extent to which each of these novel TCP congestion control variants is actually used in regulating Internet traffic, and this information, due to the complex, diverse, and decentralized nature of the Internet, is very difficult to be obtained. As a first step towards the collection of this important piece of information, we have developed a tool, presented in this paper, which allows us to identify the TCP variants employed in monitored TCP sessions. Our approach consists in passively monitoring packets and ACKs exchanged in TCP sessions, estimating the evolution of congestion window, and matching a set of features against those typical of the various TCP variants considered. Our identification tool has been validated trough emulation and real Internet experiments, showing promising results in the identification of New Reno, BIC, Cubic and Compound TCP variants.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione