Research On Service-Oriented Cloud Conferencing System Architecture And Its Key Techniques

On the rapid development of computer, network and communication technology, the IP Communication is becoming a hotspot research in computer and communication domains, which is developing from early VoIP system to the Unified Communication with diverse application. The multimedia conferencing is an IP Communication platform on network, which helps users hold conferences in Internet. The essence of the multimedia conferencing is that the VoIP is developed from end-to-end conversation to multi-party conversation, and realizes the function of the real-time transmission of voice, video and data among multi-points. Because of the limitation of network circumstance, server capability and software technology, the study and products of multimedia conferencing now still have a few problems, which infect the QoS and practicability. It is important to research on the multimedia conferencing system architecture and its key techniques to provide large scale and high quality conferencing service in Internet and mobile communication network for long-distance communication in daily life.This paper studies the characteristics of existing multimedia conferencing and finds a few weak points:small scale of conference, weak robustness of system, large delay of communication and weak reliability of media transmission. To overcome these short points, we propose a new multimedia conferencing type called the Service-Oriented Cloud Conferencing, combining cloud computing technique and service-oriented thinking. The cloud conferencing system mixes the characteristics of cloud computing technique, service-oriented thinking and multimedia conferencing. It uses service-oriented architecture to reduce coupling, enhance agility and advance expansibility. It uses cloud computing technique to enlarge the conference scale and save cost. It has the strong points of cloud computing and service-oriented architecture on the base of rich applications in multimedia conferencing. This paper also studies to provide solutions of the key techniques in cloud access method, file system, information retrieval, media transmission and process mechanism, which have scientific and practical value.The main research works of this paper are as follows:1) We study multimedia conferencing system architectures, and expatiate on the study process from centralized and distributed conferencing to cloud conferencing. This paper summarizes the author's work on centralized and distributed conferencing, and analyses their characteristics, uses and limitations. To overcome their short points and limitations, we propose the new conferencing type: Service-Oriented Cloud Conferencing. The cloud conferencing extends the two layer architecture of an open source cloud platform called Hadoop to four layers of service. This extension makes the cloud conferencing have better agility and meet the needs of different conferencing scenarios. We analyse centralized, distributed and cloud conferencing synthetically, such as size, scene, scalability, expansibility, fault tolerance, cost and delay. We find the cloud conferencing has advantages in these fields.2) We propose a method for users connecting into cloud conferencing. For the complex network environment, we choose the best cloud conferencing access point to advance communication and transmission quality. We develop Triangle Heuristic to suit the path quality evaluation, which is the theoretical basis of cloud conferencing access point chosen algorithm. We propose a dimension reduction storage method for network coordinate, which maps dimensional coordinates to leaves of B+ tree. We also propose a cloud conferencing access point chosen algorithm to improve search efficiency.3) We improve cloud conferencing file system on the base of HDFS, because of the different data types between cloud conferencing and normal cloud computing. This improvement overcomes the problems of the only NameNode's bottle-neck and weak robustness, and enhance the real-time capability. We propose an information retrieval algorithm for cloud conferencing, which is a search method for a large number of text records and documentations in cloud conferencing. We change space for time and bandwidth, on the advantage of large storage space in cloud conferencing, based on global index. We propose the key generation algorithm based on size, range and redundant condition to reduce the amount of information transmission and computing delay during information retrieval.4) The cloud conferencing has variety of scenarios and QoS demands, so we propose a new cloud conferencing media transmission strategy. This strategy uses optimal dynamic buffer, limited retransmission and flexible fault tolerance mechanisms. We can neatly demand for real-time and reliability adjustment. The Hadoop MapReduce has only one JobTracher, which has bottle-neck and weak robustness, and it can't process real-time conferencing tasks. We improve this parallel computing model MapReduce to the three step process framework called Map-Reduce-Dispatch, which meets the requirements of real-time and reliability, and can process the straggler problem.