Data Caching, Prefetching and Sharing Schemes for Multimedia Applications over the Internet

• Main Authors: Tz-Heng Hsu, 許子衡
• Other Authors: Chung-Ming Huang
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/00948390597758063049

博士 === 國立成功大學 === 資訊工程學系碩博士班 === 93 === 　Mobile Internet allows users to retrieve any information any time, any place on any device. New generation of smart wireless devices can provide an easier way to retrieve existing multimedia services for mobile users. VOIP and video conferencing are the typical popular techniques to help the communication among mobile users. The need of using multimedia services through mobile devices drives the researches of new wireless multimedia technologies.The increasing use of wireless multimedia service invokes several research issues that need to be resolved. In this dissertation, we invest in three important technical issues for distributed multimedia systems executed in the Internet: multimedia caching, prefetching, and resource　sharing. 　To speed up the transmission of multimedia objects in wireless network environments, we propose a price-based caching algorithm named GCC-DS-F-DA (GCC-DS with Frequency and Dynamic Aging). Unlike other existing caching replacement algorithms considering only the object size and transmission delay, the GCC-DS-F-DA caching algorithm considers both the requirement of (i) service and content providers and (ii) mobile users. In the GCC-DS-FDA caching algorithm, a caching profit formula is derived to estimate the reward for service providers to cache a multimedia object. By caching only those objects that have a higher gold content ratio and larger response time than others, service and content providers can get better reward and users can get better service quality. 　To decrease the retrieval latency of continuous media, we propose a user-aware prefetching scheme that uses the association rules from the data mining results. The demand of large storage space and bandwidth makes the object-based cache schemes less efficient and unsuitable in caching continuous media. The random and unpredictable user behaviors during a multimedia presentation may cause the long retrieval latency in the client-server connection. The data mining technique can provide some priority information such as the support, confidence, and association rules which can be utilized for prefetching continuous media. Thus, using the data mining technique, the proposed user-aware prefetching scheme can predict user behaviors and evaluate segments that may be accessed in near future. Performance experiments show that the proposed user-aware prefetching scheme is effective in improving the latency reduction, even for small cache sizes. 　Finally, we propose a novel P2P architecture named Shoran for interconnecting heterogeneous P2P resource sharing networks to enable connectivity and universal access of multimedia objects. Heterogeneous P2P file sharing networks need a way to collaborate and communicate with each other. Based on the approach of interconnecting heterogeneous P2P resource sharing networks, users on one P2P resource sharing network can share and search multimedia objects with other P2P resource sharing networks. Shoran provides (i) a message routing mechanism to route query messages among different P2P resource sharing networks and (ii) an XML-based uniform resource format that can ease the message exchange among heterogeneous P2P resource sharing networks. Keywords: Multimedia Proxy, Streaming Media, Multimedia Caching and Replacement Policy, Multimedia Prefetching Policy, Data Mining, Peer-to-Peer Networks, Wireless and Mobile networks, Resource sharing, Resource Discovery and Retrieval.