2005 Fiscal Year Final Research Report Summary
The Research on a High-performance Streaming System to Support Many Concurrent Requests
| Project/Area Number |
16500024
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Computer system/Network
|
|---|
| Research Institution | National University Corporation Tokyo University of Agriculture and Technology |
|---|
Principal Investigator |
YOSHIZAWA Yasufumi National University Corporation Tokyo University of Agriculture and Technology, Institute of Symbiotic Science and Technology, Division of System Information Science, Professor, 大学院・共生科学技術研究部, 教授 (10280992)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
SHINAGAWA Takahiro National University Corporation Tokyo University of Agriculture and Technology, Institute of Symbiotic Science and Technology, Division of System Information Science, Assistant Professor, 大学院・共生科学技術研究部, 助手 (40361745)
|
|---|
| Project Period (FY) |
2004 – 2005
|
| Keywords | Operating Sytem / Streaming / Cache Management |
|---|
| Research Abstract |
The purpose of the research is to develop a high-performance VoD (Video on Demand) streaming system that can support many concurrent video requests. In order to improve performance of streaming servers, reducing I/O operations is vital because streaming servers must process high-volume video contents. We achieved this by using the following three approaches.
Firstly, we developed a caching scheme that uses the memory of a client as a cache for the other clients by exploiting P2P (Peer-to-Peer) technologies. This scheme is based on the hybrid approach that uses servers together with P2P networks. By offloading caches on servers into clients, this scheme achieved reduction of both disk and network I/O operations on the server.
Secondly, we developed a scheme to distribute a video content via single multicast session to multiple clients that have different start times from each other. This scheme adopts the technique known as adaptive piggybacking : merging two unicast sessions into one multicast session by adjusting display rates of the latter request 1.3 times faster until the latter catches up the former, thereby reducing the I/O operations.
Finally, we developed a scheme to reduce the data amount of video contents without much degrading the quality. By analyzing the MPEG encoding scheme, we found that each of the three types of MPEG pictures, i.e., I-picture, P-picture, and B-picture, had different degree of impact on degradation of the quality when it was lost. The characteristic of MVs (Motion Vector) included in a picture also influences on the quality when the picture is lost. Therefore, we developed a scheme to determine pictures that give little impact on the quality of the video and reduce the data amount of video contents when network congestion occurs or fastforward is required.
|
