Sebastien Ardon

Characterizing and modelling popularity of user-generated videos

This paper develops a framework for studying the popularity dynamics of user-generated videos, presents a characterization of the popularity dynamics, and proposes a model that captures the key properties of these dynamics. We illustrate the biases that may be introduced in the analysis for some choices of the sampling technique used for collecting data; however, sampling from recently-uploaded videos provides a dataset that is seemingly unbiased. Using a dataset that tracks the views to a sample of recently-uploaded YouTube videos over the first eight months of their lifetime, we study the popularity dynamics. We find that the relative popularities of the videos within our dataset are highly non-stationary, owing primarily to large differences in the required time since upload until peak popularity is finally achieved, and secondly to popularity oscillation. We propose a model that can accurately capture the popularity dynamics of collections of recently-uploaded videos as they age, including key measures such as hot set churn statistics, and the evolution of the viewing rate and total views distributions over time.

The cost of peer discovery and searching in the Gnutella peer-to-peer file sharing protocol

A lot of attention has been focused on peer-to-peer file sharing systems. Gnutella is a fully distributed peer-to-peer protocol without the need for a central entity. This increases the reliability of the system by avoiding a single point of failure as well making it more immune to legal attack. The two main features of the Gnutella protocol discovery of peers and searching for files are implemented by passing different types of messages between the nodes of the Gnutella overlay network. Due to its fully distributed nature, Gnutella relies on flooding to route most of these messages, which immediately raises the question of cost and scalability. We study these aspects of the Gnutella protocol by means of simulation also considering the influence of the topology of the Gnutella network.

MARCH: a distributed content adaptation architecture

A novel, server-centric architecture for adapting media content to suit the operational environment for heterogeneous devices and networks is presented. The given architecture, so called MARCH, exhibits several advantages over traditional static proxy solutions. The viability of the MARCH framework has been demonstrated through a prototype implementation.

Spatio-temporal and events based analysis of topic popularity in twitter

We present the first comprehensive characterization of the diffusion of ideas on Twitter, studying more than 5.96 million topics that include both popular and less popular topics. On a data set containing approximately 10 million users and a comprehensive scraping of 196 million tweets, we perform a rigorous temporal and spatial analysis, investigating the time-evolving properties of the subgraphs formed by the users discussing each topic. We focus on two different notions of the spatial: the network topology formed by follower-following links on Twitter, and the geospatial location of the users. We investigate the effect of initiators on the popularity of topics and find that users with a high number of followers have a strong impact on topic popularity. We deduce that topics become popular when disjoint clusters of users discussing them begin to merge and form one giant component that grows to cover a significant fraction of the network. Our geospatial analysis shows that highly popular topics are those that cross regional boundaries aggressively.

First impressions on the state of cellular data connectivity in India

Cellular penetration in India has grown tremendously in recent years and provides an opportunity to bridge the digital divide. However, there is little understanding of the state of cellular data connectivity in India. In this paper, we present first impressions on cellular data network performance in India. We present a measurement framework designed specifically for remote deployments and intermittent connectivity. Using this framework we evaluate three GSM based and one CDMA based cellular service providers through active measurements conducted at five rural, one semi-urban, and one urban locations. Through analysis of about 450 hours of measurement data collected over a 3-month period, we present the throughput and latency performance of cellular service providers and provide insights into the architecture of the service provider networks. Our analysis reveals aspects in cellular network design that interfere with standard protocols such as TCP, and suggests ways to improve performance.

Geolocating IP addresses in cellular data networks

Smartphones connected to cellular networks are increasingly being used to access Internet-based services. Using data collected from smartphones running a popular location-based application, we examine IP address allocation in cellular data networks, with emphasis on understanding the applicability of IP-based geolocation techniques. Our dataset has GPS-based location data for approximately 29,000 cellular network assigned IP addresses in 50 different countries. Using this dataset, we provide insights into the global deployment of cellular networks. For instance, we find that Network Address Translation (NAT) is commonplace in cellular networks. We also find several instances of service differentiation with operators assigning public IP addresses to some devices and private IP addresses to other devices. We also evaluate the error of geolocation databases when determining the position of the smartphones, and find that the error is 100km or more for approximately 70% of our measurements. Further, there is potential for errors at the scale of inter-country and inter-continent distances. We believe this dataset may be of value to the research community, and provide a subset of the dataset to the community.

Toward efficient on-demand streaming with bittorrent

This paper considers the problem of adapting the BitTorrent protocol for on-demand streaming. BitTorrent is a popular peer-to-peer file sharing protocol that efficiently accommodates a large number of requests for file downloads. Two components of the protocol, namely the rarest-first piece selection policy and the tit-for-tat algorithm for peer selection, are acknowledged to contribute toward the protocols efficiency with respect to time to download files and its resilience to free riders. Rarest-first piece selection, however, is not suitable for on-demand streaming. In this paper, we present a new adaptive window-based piece selection policy that balances the need for piece diversity, which is provided by the rarest-first algorithm, with the necessity of in-order piece retrieval. We also show that this simple modification to the piece selection policy allows the system to be efficient with respect to utilization of available upload capacity of participating peers, and does not break the tit-for-tat incentive scheme which provides resilience to free riders.

Catch-up TV recommendations: show old favourites and find new ones

Web-based catch-up TV has revolutionised watching habits as it provides users the opportunity to watch programs at their preferred time and place, using a variety of devices. With the increasing offer of TV content, there is an emergent need for personalised recommendation solutions, which help users to select programs of interest. In this work, we study the watching patterns of users of an Australian nation-wide catch-up TV service provider and develop a suite of approaches for a catch-up recommendation scenario. We evaluate these approaches using a new large-scale dataset gathered by the Web-based catch-up portal deployed by the provider. The evaluation allows us to compare the performance of several recommenders that address the discovery of both TV programs already watched by users and new programs that users may find relevant.

Characterizing and Predicting Viral-and-Popular Video Content

The proliferation of online video content has triggered numerous works on its evolution and popularity, as well as on the effect of social sharing on content propagation. In this paper, we focus on the observable dependencies between the virality of video content on a micro-blogging social network (in this case, Twitter) and the popularity of such content on a video distribution service (YouTube). To this end, we collected and analysed a corpus of Twitter posts containing links to YouTube clips and the corresponding video meta-data from YouTube. Our analysis highlights the unique properties of content that is both popular and viral, which allows such content to attract high number of views on YouTube and achieve fast propagation on Twitter. With this in mind, we proceed to the predictions of popular-and-viral clips and propose a framework that can, with high degree of accuracy and low amount of training data, predict videos that are likely to be popular, viral, and both. The key contribution of our work is the focus on cross-system dynamics between YouTube and Twitter. We conjecture and validate that cross-system prediction of both popularity and virality of videos is feasible, and can be performed with a reasonably high degree of accuracy. One of our key findings is that YouTube features capturing user engagement, have strong virality prediction capabilities. This findings allows to solely rely on data extracted from a video sharing service to predict popularity and virality aspects of videos.

EMO: A statistical encounter-based mobility model for simulating delay tolerant networks

We propose EMO, a model to evaluate delay tolerant networks (DTN) and opportunistic systems, which focuses on simulating encounter events between mobile radios, rather than node locations as done in existing models and simulators. Our approach introduces a more accurate simulation of DTNs on the main system timescale (the encounter timescale), while trading off some accuracy at the bit-level, through an abstraction of radio propagation simulation. To design EMO, we extract and characterize the necessary parameters from experimental data and propose a method to generate synthetic node encounter traces based on this characterization. The output of the model is validated using hold-out cross-validation method. Our validation results indicate that EMO is able to maintain the statistical properties of experimental data over a wide range of time (simulation duration) and space (number of nodes) scales, with mean square errors of less than 3% for the main system parameters.