A Novel Transport Service Model with Partial Reliability and Causal Ordering for Real-Time Applications

• Main Authors: Chao-Cheng Wen, 溫照成
• Other Authors: Kim-Joan Chen
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/5pest2

博士 === 國立中正大學 === 電機工程所 === 94 === In the past twenty or thirty years, the network users increase exponentially due to the widely deployed networks, especially the Internet, which changes the human life in many ways. Many social activities in people life change to the computer networks, which motivates more emerging applications to be created. Unlike the traditional Internet applications, that are mainly file transfer and email transferfootnote{From the viewpoint of the transport layer, Email can be viewed as small file transfer application.}, these emerging applications are real-time interactive and usually group-based, multipoint communications, for example, Grid Computing, Web Service, P2P tchnology, messenger service, and video conferencing. The communication model of the applications have changed, and the transport requirements of the application nowadays are also different from that several-tens years before. In the meantime, the evolution of the networking technology stimulates the changes of the Internet. Although the Internet is the universal network platform in today networks, however, different network service providers use different internetworking technology, e.g., ATM networks, MPLS networks, to support their special demands for control and management purposes. Therefore, the transfer capability and quality of the Internet are different than before, and it is a heterogeneous networks now. The quality of the network platform changes, too. However, the transport service, which is layered between the applications and the networks and responsible for the adaption of the network transfer capability and the application requirements, doesn't change as its upper- and lower layers change. Many real-time applications still use the traditional TCP and UDP transport protocols, which doesn't provide new transport functionalities that the applications need, so that, the performance of the real-time applications may be poor and the transport service model is impertinent regarding these applications. In this dissertation, we propose a novel transport service for the future applications, which integrates the most important transport functionalities into one transport service, including the guarantees of message causality, timely and reliable delivery. The proposed transport service model, called Multipoint Integrated Transport Service Model (MITSM), considers the quality variance of the network transfers, e.g., ene-to-end delays, such that the novel transport service can adapt the quality variance of the network transfers to the requirements of the applications. In the MITSM, we believe, the applications will adopt multipoint communications model, which means that, every end process play both of sender and receiver roles. The sent messages and the received messages in one process are correlated in order, which also have order dependence on the messages sent from the other to another. Besides, these multipoint applications usually requires real-time delivery, which means that the application entities wish to get the messages from the others in time. Furthermore, the applications are often deployed over the global networks sparsely, in which the reliability of the sent messages is an important issue. In general, the applications require the transport layer to support partial reliability, not total reliability as the traditional traditional transport protocol provides. Traditional transport protocol, such as TCP, use complicated retransmission mechanism to protect and recover the lost messages, which may degrade the throughput performance of the real-time applications. Therefore, the concept of partial reliability is popular in modern transport services, which can get the balance between real-time and reliable delivery. The MITSM uses causal ordering to protect the message causality, meanwhile, it concerns about the three requirements of real-time and reliable delivery with respect to the applications. In other words, the applications get the transport guarantees of the three features. In this dissertation, we also explain how to realize the functionalities, mentioned in the MITSM, for the three transport requirements. Furthermore, we implement a prototype of such transport service, called Real-Time Transport Service for Multipoint Applications (RTSM), and take experiments in the Internet for performance analysis as well as the insight observations of the transport service, including the trade-off between the three requirements. We believe that our works can provide a more flexible and sophisticated transport service to the real-time applications.