Sébastien Monnet

Scattering and Placing Data Replicas to Enhance Long-Term Durability

Distributed storage systems have to ensure data availability and durability despite the occurrence of failures. To do so, many of them rely on replication mechanisms. We show that the layout of the data block copies on the nodes, chiefly the way the copies are scattered, has a major impact on the reparation speed and thus on the data loss ratio. In this paper, we propose an approach that provides the ability: (i) to finely tune the proportion of common content stored by the nodes, and (ii) to control the storage load distribution while creating new data block copies. We propose a simulation model that allows us to present a long-term study of the impact of the data block copies layout and the systems storage load on the data loss ratio.

MOve: Design of An Application-Malleable Overlay

Peer-to-peer overlays allow distributed applications to work in a wide-area, scalable, and fault-tolerant manner. However, most structured and unstructured overlays present in literature today are inflexible from the application viewpoint. In other words, the application has no control over the structure of the overlay itself. This paper proposes the concept of an application-malleable overlay, and the design of the first malleable overlay which we call MOve. In MOve, the communication characteristics of the distributed application using the overlay can influence the overlays structure itself, with the twin goals of (1) optimizing the application performance by adapting the overlay, while also (2) retaining the large scale and fault tolerance of the overlay approach. The influence could either be explicitly specified by the application or implicitly gleaned by our algorithms. Besides neighbor list membership management, MOve also contains algorithms for resource discovery, update propagation, and churn-resistance. The emergent behavior of the implicit mechanisms used in MOve manifests in the following way: when application communication is low, most overlay links keep their default configuration; however, as application communication characteristics become more evident, the overlay gracefully adapts itself to the application

MOve: Design and Evaluation of a Malleable Overlay for Group-Based Applications

While peer-to-peer overlays allow distributed applications to scale and tolerate failures, most structured and unstructured overlays in literature today are inflexible from the application viewpoint. The application thus has no first-class control on the overlay structure. This paper proposes the concept of an application-malleable overlay, and the design of the first malleable overlay which we call MOve. MOve is targeted at group- based applications, e.g., collaborative applications. In MOve, the communication characteristics of the distributed application using the overlay can influence the overlays structure itself, with the twin goals of (1) optimizing the application performance by adapting the overlay, while also (2) retaining the large scale and fault tolerance of the overlay approach. Besides neighbor list membership management, MOve also contains algorithms for resource discovery, update propagation, and churn-resistance. The emergent behavior of the implicit mechanisms used in MOve manifests as follows: when application communication is low, most overlay links keep their default configuration; however, as application communication characteristics become more evident, the overlay gracefully adapts itself to the application. We validate MOve using simulations with group sizes that are fixed, uniform, exponential and PlanetLab-based (slices), as well as churn traces and two sample management-based applications.

SDN-based Wi-Fi Direct Clustering for Cloud Access in Campus Networks

Mobile cloud is changing the way to enroll teaching activities in a university campus. Lectures and lab sessions can be carried out directly from tablets in a classroom by accessing a server in the cloud. In this paper, we address the problem of high-density cloud access with wireless devices in campus networks. We propose to use Wi-Fi direct clustering to solve the problem of quality of service (QoS) degradation when a high number of wireless devices want to access a content in the cloud at the same time. A centralized software-defined network controller is used in our proposed architecture to capture the network state and organize the Wi-Fi Direct groups. The optimized number of clusters can be calculated in function of the number of devices in the room. By simulations, we show that we can provide a better QoS in terms of download time and application’s throughput by reducing the interference in this dense wireless network environment.

Analysis of a Stochastic Model of Replication in Large Distributed Storage Systems: A Mean-Field Approach

Distributed storage systems such as Hadoop File System or Google File System (GFS) ensure data availability and durability using replication. Persistence is achieved by replicating the same data block on several nodes, and ensuring that a minimum number of copies are available on the system at any time. Whenever the contents of a node are lost, for instance due to a hard disk crash, the system regenerates the data blocks stored before the failure by transferring them from the remaining replicas. This paper is focused on the analysis of the efficiency of replication mechanism that determines the location of the copies of a given file at some server. The variability of the loads of the nodes of the network is investigated for several policies. Three replication mechanisms are tested against simulations in the context of a real implementation of a such a system: Random, Least Loaded and Power of Choice. The simulations show that some of these policies may lead to quite unbalanced situations: if ß is the average number of copies per node it turns out that, at equilibrium, the load of the nodes may exhibit a high variability. It is shown in this paper that a simple variant of a power of choice type algorithm has a striking effect on the loads of the nodes: at equilibrium, the distribution of the load of a node has a bounded support, most of nodes have a load less than 2ß which is an interesting property for the design of the storage space of these systems. Stochastic models are introduced and investigated to explain this interesting phenomenon.

EDWiN: Leveraging Device-to-Device Communications for Efficient Data Dissemination over Wi-Fi Networks

An emerging usage is to rely on mobile devices (Smartphones or tablets) for large-scale events. They can be used for many applications like live voting or chatting, but also to access all the data related to an event. However, in such case, handling mobile devices trying to access data simultaneously is difficult. A Wi-Fi access point can only handle a limited amount of devices. Current solutions, relying on pre-loading data on the devices or over-sizing the network equipments are not satisfying and are not even always possible. We propose an approach that leverages the capability of mobile devices to interact directly through device-to-device (D2D) communications. Our solution can be tuned to choose the right level of parallelization to cope with radio interferences, it also provides the ability to adjust the trade-off between efficiency and energy consumption. We evaluate our approach using a discrete event simulator. The simulation results show that our approach using D2D communications brings a 30% gains.

Analysis of a Stochastic Model of Replication in Large Distributed Storage Systems

Distributed storage systems such as Hadoop File System or Google File System (GFS) ensure data availability and durability using replication. Persistence is achieved by replicating the same data block on several nodes, and ensuring that a minimum number of copies are available on the system at any time. Whenever the contents of a node are lost, for instance due to a hard disk crash, the system regenerates the data blocks stored before the failure by transferring them from the remaining replicas. This paper is focused on the analysis of the efficiency of replication mechanism that determines the location of the copies of a given file at some server. The variability of the loads of the nodes of the network is investigated for several policies. Three replication mechanisms are tested against simulations in the context of a real implementation of a such a system: Random, Least Loaded and Power of Choice. The simulations show that some of these policies may lead to quite unbalanced situations: if β is the average number of copies per node it turns out that, at equilibrium, the load of the nodes may exhibit a high variability. It is shown in this paper that a simple variant of a power of choice type algorithm has a striking effect on the loads of the nodes: at equilibrium, the distribution of the load of a node has a bounded support, most of nodes have a load less than 2β which is an interesting property for the design of the storage space of these systems. Stochastic models are introduced and investigated to explain this interesting phenomenon.

étude des services de matchmaking dans les jeux multijoueurs en ligne. Récupérer des traces d’utilisateurs pour améliorer l’expérience de jeu

L’acquisition et l’analyse de traces d’utilisation reelles issues de jeux video permettent d’acquerir une connaissance approfondie des habitudes, des comportements et des attentes des joueurs. Cela constitue une etape cruciale pour la conception de services logiciels de qualite pour les jeux video. Toutefois, les acteurs de cette industrie protegent jalousement de telles don- nees afin de dissuader une eventuelle reutilisation par la concurrence et pour eviter la triche. Nous avons contourne ce probleme en collectant, sur le serveur d’un jeu en vogue (League of Legends), des donnees publiques representant plus de 28 millions de sessions. Dans cet article, nous presentons notre base de donnees disponible librement en ligne. Nous detaillons ensuite l’analyse et les conclusions que nous tirons de ces donnees concernant les exigences d’un service essentiel pour les jeux multi-joueurs en ligne: le matchmaking, ou mise en relation de joueurs.

POPS: A popularity-aware live streaming service

Live streaming has become very popular. Many systems, such as justin.tv, have emerged. They aim to collect user live-streams and serve them to the viewers using broadcasting servers. However, the huge variation in the total number of viewers and the great heterogeneity among streams popularity generally implies over-provisioning, leading to an important resource waste. In this paper, we show that there is a trade-off between the number of servers involved to broadcast the streams and the bandwidth usage among the servers. We also stress the importance to predict streams popularity in order to efficiently place them on the servers. We propose POPS: a live streaming service using popularity predictions to map live-streams on the servers.