Logo image
Distributed Storage and Delivery for Popular Videos using LT Codes
Thesis

Distributed Storage and Delivery for Popular Videos using LT Codes

Yen, Chia-Cheng
Masters, 國立清華大學, 資訊工程學系
2013

Abstract

分散式儲存 分散式解碼 資料配置 資料傳輸 可適性視訊編碼 盧比變換碼 無線區域網路 Distributed storage Distributed decoding Data allocation Data delivery Scalable Video Coding LT codes WLAN
The application of video streaming is expected to shift to mobile broadband as soon as the High efficiency WLAN (HEW), Beyond 4G technologies and related devices become large-scale and dense deployed. The motivation of this thesis is to study distributed storage and delivery technologies for streaming popular videos over such ultra-dense error-prone wireless environment. The H.264 SVC scalable source coding and the LT codes of rateless channel coding are both considered and integrated to provide reliable and scalable video services wirelessly. To serve popular requests with broadband and high-speed requirements, one nearby node (server) immediately decode its own LT encoded data once a request coming and broadcast them to the clients and the neighboring nodes to proceed the undone decoding process in parallel. Thus, the clients can receive data from multiple nodes to achieve the goal of bandwidth aggregation. Two storage allocation schemes are considered to balance the storage space, transmission bandwidth and fault tolerance requirements. Besides, once the decoding process being completed, the repairing process is ignited to recover the possible node failures. In the storage allocation process, the video segments are applied SVC encoder first, and coded by LT codes or the Expanding Window LT codes. Then, the LT encoded data are randomly distributed among nodes. The data proportion of SVC layers is tuned to let the base layer being completely decoded first. In the delivery process, a distributed LT decoder is proposed to decode and transmit in parallel, where each node passing its own self-generated degree one data with neighbors by broadcasting. Finally, the delivery with caching mechanism is proposed to serve hot requests effectively. In the ultra-dense scenarios, our experiments show that each request can be served with at least base layer and with one more enhancement layer in average under communication loss rate ranging 1%~10% and possible node failures. Keywords: Distributed storage, Distributed decoding, Data allocation, Data delivery, Scalable Video Coding, LT codes, WLAN.

Metrics

1 Record Views

Details

Logo image