The Internet has recently been evolving from homogeneous congestion control to heterogeneous congestion control. Five years ago, Internet traffic was mainly controlled by the standard TCP AIMD algorithm, whereas Internet traffic is now controlled by many different congestion control algorithms. However, there is very little work on the performance and stability study of the Internet with heterogeneous congestion control. One fundamental reason is the lack of the deployment information of different congestion control algorithms. Motivated by the increasing number of public, private, and proprietary congestion control algorithms deployed in the Internet and the lack of the information about the heterogeneous Internet, the goals of this project are to 1) design methodologies and tools for identifying the congestion control algorithms and related parameters of Internet nodes and flows, and 2) conduct large scale comprehensive measurements for identifying and evaluating the congestion control algorithms in the heterogeneous Internet. As an analogy, if we consider the Internet as a country, an Internet node as a house and a congestion control algorithm running at a node as a person living at a house, this project is to conduct a population census in the country of the Internet. Just like the population census is vital for the study and planning of the society, the proposed congestion control census is vital for the study and planning of the Internet. The expected results of this project include methodologies and tools for identifying congestion control algorithms, and a congestion control census database containing measurement results.
Congestion control is one of the fundamental components of the Internet, since it largely determines the performance and stability of the Internet. Several years ago, the Internet was still a homogenous system in the sense that most Internet traffic was controlled by the same congestion control algorithm --- the standard TCP Additive-Increase-Multiplicative-Decrease (AIMD) algorithm. However, the Internet has recently been evolving into a heterogeneous system in that the Internet traffic is now controlled by a large number of different congestion control algorithms. Most of these algorithms, such as CTCP, BIC, CUBIC, HSTCP, HTCP, and FAST, have recently been developed for achieving better performance than the traditional AIMD in the emerging high-speed long-distance networks. The performance and stability of the homogenous Internet has been extensively studied in the past using methods such as network simulation, Internet experiments, control theory, and optimization theory. However, there are very few works that investigate the performance and stability of the heterogeneous Internet. One fundamental reason is the lack of the information about the heterogeneous Internet. For example, we do not even know the answers to the following basic question: What percentage of Internet users are using a specific congestion control algorithm? Motivated by the increasing number of public, private, and proprietary congestion control algorithms deployed in the Internet and the lack of the information about the heterogeneous Internet, the goals of this Internet Congestion Control Census project are to 1) design methodologies and tools for identifying the congestion control algorithms and related parameters of Internet nodes and flows, and 2) conduct large scale comprehensive measurements for identifying and evaluating the congestion control algorithms in the heterogeneous Internet. As an analogy, if we consider the Internet as a country, an Internet node as a house, and a congestion control algorithm running at a node as a person living at a house, this project is to conduct a population census in the country of the Internet. Just like the population census is vital for the study and planning of the society, the proposed congestion control census is vital for the study and planning of the Internet. During this project, we have designed and developed a method, called CAAI, to identify the TCP congestion control algorithm of a web server. CAAI can identify all default TCP algorithms and most non-default TCP algorithms of major operating system families. For the first time, we have conducted a large scale of measurement of the deployment information of TCP congestion control algorithms of over 60,000 popular web servers. Our measurement results show that the majority of TCP flows are not controlled by the traditional AIMD anymore, and that the Internet congestion control has already changed from homogeneous to highly heterogeneous. One undergraduate student, one female graduate student, and more than five other graduate students were involved this project. They participated in designing the measurement methods, developing the software, conducting the Internet measurements, and writing the research papers.
