Distributed delivery system for time-shifted streaming systems

Abstract

To cite this version:Yaning Liu, Gwendal Simon. Distributed delivery system for time-shifted streaming systems. Abstract-In live streaming systems (IPTV, life-stream services, etc.), an attractive feature consists in allowing users to access past portions of the stream. This is called a time-shifted streaming system. We address in this paper the design of a large-scale delivery system for a time-shifted streaming application. We highlight the challenging characteristics of time-shifted applications that prevent known delivery systems to be used. Then, we describe the turntable structure, the first structure that has been specifically designed to cope with the properties of time-shifted systems. A set of preliminary simulations confirm the interest for this structure. I. INTRODUCTIONThe delivery of television over the Internet (IPTV) is expected to offer viewers new ways to enjoy TV content. One of the most promising services, often called catch-up TV or time-shifted TV, consists in allowing viewers to watch their favorite broadcast TV programs within an expanded time window. Let's say that a program is normally broadcasted from a given time t. In a catch-up TV, this program is made available for viewing at any time from t to t + δ hours where δ can be excessively long (several weeks). In this context, a viewer is also able to surf the TV content history using pause, rewind or fast forward commands, hence he/she can switch from a live experience to a shifted one.Today, to enjoy catch-up TV requires to record the stream on a Digital Video Recorder (DVR) connected to Internet. Of course, this is unacceptable for TV providers, which would like to control the delivery of their content. However, building a large-scale time-shifted streaming service is not trivial. Indeed, the disk-based servers that are currently used in ondemand video services (VoD) have not been designed for concurrent read and write operations. In particular, a VoD server can not massively ingest content. Moreover, delivery systems for IPTV can not be utilized because, contrarily to live streaming systems, time-shifted systems can not directly use group communication techniques like multicast protocols, for the reason that clients require distinct portions of the stream. Other obvious differences include the length of a catch-up TV stream, which can be several orders of magnitude longer than a typical movie in VoD, and the dynamicity of chunk request. Contrarily to VoD, the popularity of every chunk is variable in catch-up TV.A few papers have recently addressed the VCR problem in peer-to-peer VoD systems [1,2], but no previous work has assumed that VCR is so massively employed by user.

Time-Shifted TV in Content Centric Networks: The Case for Cooperative In-Network Caching

“…A series of recent works has explored CDN-based and peerto-peer approaches for time-shifted TV [9]- [14]. However, none of these solutions takes ISPs' behalf into account.…”

“…Hence, it may happen that the video is downloaded from one or several distant peers. In our previous works [14], we have addressed the problem of guarantying that all past chunks are correctly kept in a peer-to-peer system.…”

“…We have used the synthetic traces generated in [14] to emulate the behavior of users of time-shifted services. We have also implemented the ISP topology measured by Rocketfuel [17].…”

“…We use the same synthetic model as [14] for modeling the behavior of users of time-shifted services. This model is based on two measurement studies conducted in 2008 and 2009 [8,18].…”

Peer-Assisted Time-Shifted Streaming Systems: Design and Promises

“…These works however do not tackle the scalability issue met by time-shifted streaming systems. A peer-to-peer architecture for time-shifted streaming systems has been addressed in [7][8][9]. These systems are based on a predefined achitechture, such as Distributed Hash Table (DHT) or turntable.…”