Abstract: The authors propose a multiple server parallel transport approach using Luby-Transform (LT) codes. With LT codes, a client could download the same segment from multiple servers without considering data segmentation and server scheduling issue.Thus, all heterogeneous servers are combined as a virtual server with higher bandwidth. To avoid transmitting redundant data when waiting for ACK messages (ACK overhead) after requesting data from each server, an algorithm based on traffic prediction models is proposed, to be more specific, a stochastic model is constructed to predict the amount of data required from each server for ensuring success of decoding precedure. Compared with existing approaches, the experimental results show that the proposed approach obtains less transmission overhead and fewer number of HTTP requests. Besides, it achieves better video quality and higher robustness to fluctuated bandwidth.

Key words: HTTP streaming, multi-Server, LT codes

摘要： 提出一种基于Luby-Transform (LT)编码的多服务器HTTP 流媒体传输方法。在服务器端, 通过对数据进行LT 编码, 多台异构服务器被等效为一台具有更大带宽的虚拟服务器, 客户端可以从多台服务器并行下载任意编码后的数据片段, 而不用考虑数据分块和服务器调度等问题。为了减少传输过程中由于等待ACK 消息而产生的冗余数据(ACK 开销), 提出一种基于流量模型的数据请求算法, 构建网络随机模型, 预测每台服务器需要传输的时间。实验结果表明, 与传统的并发传输方法相比, 所提方法可以大大降低传输开销, 减少HTTP 请求次数, 获得更好的视频质量, 对波动带宽有更好的鲁棒性。

关键词: HTTP 流媒体, 多服务器, LT 编码