통신망 설계 및 분석
Statistical Research Center for Complex Systems
지원기간 : 2000 - 2008
지원기관 : 과학기술부 우수연구센터 육성사업단(SRC/ERC)
 

Research Area
 This research area specializes in modeling the traffic of various forms. Network traffic consists of many different types of applications such as the famous HTTP, FTP, and emails. However as the network technology and the physical layer of the architecture advances in rapid and significant speed, new applications such as multimedia, voice, game, and download of large files frequently occupies the network traffic. Each of the applications behaves in rather differently, and has different requirements. Voice traffic has to be sent immediately, meaning that it does not tolerate latency, but the sizes of the packets are very small. Another example is multimedia contents. The application requires high bandwidth and also does not accept latent packets for quality of service purposes. Thus modeling the network traffic which consists of many different applications is indeed an important issue.

 Many of the traffic modeling and analyzing projects are done theoretically. Because acquiring process for the live traffic traces are extremely difficult for security reasons as well as privacy reasons. Most of the traces, which an institute proclaims to have, are traffic traces of few departments of university, or a building of a company. Some of the traces from Sprint Co. makes available of their traffic traces for research issues, however it is not easy to have access to that data since they have special contacts only to some of major Internet measurement centers and laboratory. Some of other institutes lets researchers to access their traffic data, but as it is tangled with many privacy laws and valuable company information, they transform the traffic traces to sanitized data, which is not enough for analysis in many cases. DMC Lab. is fortunate to have the live traffic trace data, which consumes about 1.5 Tbytes of packets with payload with granularity of 0.1 micro sec. It is captured from live and up-and-running Internet Exchange point with about 155Mbits/sec of bandwidth.



Figure 1. Modeling of Network Traffic
 Having such amount of valuable traffic data, multidimensional analysis and through characterization is possible. UDP and TCP behavior is decided by its protocol difference. UDP utilizes connectionless oriented service, and the TCP uses connection oriented service, because UDP does not use three way handshakes to connect the server, or has congestion control mechanism. People have modeled TCP with self similarity, markovian process, fractal models, which is most of the time deals with one dimensional view of the traffic trace. In this project DMC Lab. performs multi dimensional analysis of network traffic capturing the bandwidth process and packet count process.
 
Research Goal
   1. Constructs model for I/O traffic of communication network and researches a performance evaluation methodology for it, which is used to develop a methods of optimizing communication network, controlling traffic
   2. Develops a effective algorithm to produce reliability for the communication network structure, which is applied to the reliability distribution of the component of the network
 
Expected Effects
   1. Constructs a performance analysis methods. Improves a quality of evaluation for servies. Improves the technology of controlling traffic of the network. Makes advanced service through effective control. Contributes to early construct the next generation communication network through effective controlling for heavy traffic.
   2. Maximizes the satisfaction of the communication network users. Enables the network administrator to easily manage the network. Provides the useful information in network restructuring.
 
Supporting Institute: 과학기술부 우수연구센터 육성사업단(SRC/ERC)
Duration of the Research: 2000 - 2008
 
Papers
Research Personnel
  • Junghyun Kim junghyun<>ece.hanyang.ac.kr
  • Seongjin Lee james<>ece.hanyang.ac.kr