This two days workshop addressed distributed storage technologies and issues (e.g., consistency, redundancy, placement, privacy, security). The workshop had been held in Technicolor Rennes, 1 Avenue de Belle-Fontaine, Cesson-Sevigné.
The summary of the presentations follows. You can download slides and technical reports describing the work if they are available.
Session 1: P2P Distributed Backup and Sharing
The Wuala file sharing and backup system, Thomas Mager (EURECOM) – Slides
Efficient P2P backup through buffering at the edge, Alex Van Kempen (Technicolor) – Slides
Performance Analysis of P2P Storage Systems, Remigiusz Modrzejewski (INRIA) – Slides
Session 2: Coding for Storage Systems
Overview of Regenerating Codes, Nicolas Le Scouarnec (Technicolor) – Slides
Exact Regenerating Codes on Hierarchical Codes, Ernst Biersack (EURECOM) – Slides
Session 3: Managing redundancy in distributed systems
Exploiting Rateless Coding in Structured Overlays to Achieve Data Persistence, Emmanuelle Anceaume (CNRS) – Slides
Availability-based methods for distributed storage, Erwan Le Merrer (Technicolor)
Session 4: Managing User Data and Enabling Sharing
Distributed Content Sharing and Backup using Social Information, Jin Jiang (Polytechnico di Torino)
Pretty Private Group Management, Olivier Heen (Technicolor)
Home Networking as a Distributed File System View, Serge Defrance (Technicolor) – Slides
Demo Session (Unified Home Storage), Jean Le Roux (Technicolor)
Jin Jiang, Polytechnico di Torino
Distributed Content Sharing and Backup using Social Information
In this talk, we consider the case that every household is equipped with a home gateway that stores and manages the data collected by the home users. To combine the benefit of content sharing and backup, we propose an efficient backup scheme that hinges upon gateway interactions exploiting the users’ social networking information. We formulate this problem as a Budgeted Maximum Coverage (BMC) problem and examine its solution under a synthetic social network scenario. We then propose some distributed implementations of our scheme based on heuristics that approximate the optimum distribution and discuss their merits.
More Information: Socially-aware Gateway-based Content Sharing and Backup, Jin Jiang and Claudio Casetti, SIGCOMM HomeNets 2011
Ernst Biersack, EURECOM, Nice
Exact Regenerating Codes on Hierarchical Codes (joint work with H. Zheng)
Peer to peer backup systems store data on “unreliable” peers that can leave the system at any moment. In this case, the only way to assure durability of the data is to add redundancy using either replication or erasure codes. Erasure codes are able to provide the same reliability as replication requiring much less storage space. Erasure coding breaks the data into blocks that are encoded and then stored on different nodes. However, when storage nodes permanently abandon the system, new redundant blocks must be created, which is referred to as repair. For “classical” erasure codes, generating a new block requires the transmission of k blocks over the network, resulting in a high repair traffic. Recently, two new classes of erasure codes, Regenerating Codes and Hierarchical Codes, have been proposed that significantly reduce the repair traffic: Regenerating Codes reduce the amount of data uploaded by each peer involved in the repair, while Hierarchical Codes reduce the number of nodes participating in the repair. In this paper we propose to combine these two codes to devise a new class of erasure codes called ER-Hierarchical Codes that combine the advantages of both.
Slides of the talk
Thomas Mager, EURECOM, Nice
The Wuala file sharing and backup system
In 2008, the free distributed social storage service Wuala was launched. Based on a solid server infrastructure combined with innovative Peer-to-Peer technology, it leverages system resources of participating users to enhance service and save costs. Since the team of Wuala is a pioneer in building a real world application in this field, interest arouses in their released implementation. It is of particular interest how they cope with requirements for stored files concerning reliability, durability and availability on mainly unreliable storage. Therefore, this work analyses the Wuala software client by primarily observing the connectivity to other peers and servers provided by the team of Wuala. It reveals how files are encoded before their data is distributed over the network. Further it unveils that the system is, at least nowadays, relying significantly stronger on servers than claimed in announcements of the Wuala team. Indeed, Wuala would not be functional without their servers, whereas peers are mainly responsible for serving frequently accessed files to save load on servers.
Slides of the talk
Nicolas Le Scouarnec, Technicolor, Rennes
Overview of Regenerating Codes
Regenerating codes are the application of network coding to distributed storage system by Dimakis et al.. Studies in this area focus on the problem of repairing efficiently a system relying on code-based redundancy. By casting the repair problem into a multicast problem and using network coding, regenerating codes have allowed both the characterization of optimal tradeoffs between storage and repair bandwidth and the definition of new codes offering good performances. The initial works have been followed by studies on the definition of practical regenerating codes and studies of slightly different models. In this talk, we will give a brief overview of research around the use of network coding for distributed storage, and we will present our contributions concerning the repair of multiple failures.
Slides of the talk
More information: Repairing Multiple Failures with Coordinated and Adaptive Regenerating Codes, A.-M. Kermarrec, Nicolas Le Scouarnec and Gilles Straub , NetCod 2011
Alexandre Van Kempen, Technicolor, Rennes
Efficient peer-to-peer backup services through buffering at the edge
The availability of end devices of peer-to-peer storage and backup systems has been shown critical for usability and for system reliability in practice. This has led to the adoption of hybrid architectures composed of both peers and servers. Such architectures mask the instability of peers thus approaching the performances of client-server systems while providing scalability at a low cost. In this presentation, we advocate the replacement of such servers by a cloud of residential gateways, as they are already present in users’ homes, thus pushing the required stable components at the edge of the network. In our gateway-assisted system, gateways act as buffers between peers, compensating for their intrinsic instability. This enables to offload backup tasks quickly from the user’s machine to the gateway, while significantly lowering the retrieval time of backed up data. We evaluate our proposal using real world traces including existing traces from Skype and Jabber as well as a trace of residential gateways for availability, and a residential broadband trace for bandwidth. Results show that the time required to backup data in the network is comparable to a server-assisted approach, while substantially improving the time to restore data, which drops from a few days to a few hours. As gateways are becoming increasingly powerful in order to enable new services, we expect such a proposal to be leveraged on a short term basis.
Slides of the talk
More information: Efficient peer-to-peer backup services through buffering at the edge, Serge Defrance, Anne-Marie Kermarrec, Erwan Le Merrer, Nicolas Le Scouarnec, Gilles Straub and Alexandre Van Kempen, P2P 2011
Erwan Le Merrer, Technicolor, Rennes
Availability-based methods for distributed storage systems
Distributed storage systems rely heavily on replication to ensure data availability as well as durability. In networked systems subject to intermittent node unavailability, replicas need to be maintained (i.e. replicated and/or relocated upon failure). Repairs are well-known to be extremely bandwidth-consuming and it has been shown that, without care, they may significantly congest the system. In this presentation, we propose an approach to replica management accounting for nodes heterogeneity with respect to availability. We show that by using the availability history of nodes, the performance of two important faces of distributed storage (replica placement and repair) can be significantly improved. Replica placement is achieved based on complementary nodes with respect to nodes availability, improving the overall data availability. Repairs can be scheduled thanks to an adaptive per-node timeout according to node availability, so as to decrease the number of repairs while reaching comparable availability. We propose practical heuristics for those two issues. We evaluate our approach through extensive simulations based on real and well-known availability traces. Results clearly show the benefits of our approach with regards to the critical trade-off between data availability, load-balancing and bandwidth consumption.
More information: Availability-based methods for distributed storage systems, Anne-Marie Kermarrec, Erwan Le Merrer, Gilles Straub and Alexandre Van Kempen, Technical Report, HAL
Olivier Heen, Technicolor (Security Lab), Rennes
Pretty Private Group Management
Group management is a fundamental building block of today’s Internet applications. Mailing lists, chat systems, collaborative document edition systems but also online social networks such as Facebook and Twitter use of group management systems. These applications must scale to hundreds of millions of users and at the same time offer group security and privacy. However, group management systems often rely on a central authority that manages and controls the system’s infrastructure and data. Thus, personal user related to groups becomes accessible to the central authority. We believe that only a distributed system with no central authority is able to resolve all scalability, security and privacy issues at once. We propose a completely distributed approach for group management requiring no central authority, based on Distributed Hash Tables. As no enrollment to a central authority is required, the system can be leveraged by various applications that require group management. We partly address the specific security and privacy problems that are inherently introduced by removing the central authority. The system offers a set of security and privacy properties against adversaries that have compromised nodes of the system. We provide a formal validation of security properties of the protocol using AVISPA and show to what extent the proposed system meets privacy objectives.
Emmanuelle Anceaume, CNRS, Rennes
Exploiting Rateless Coding in Structured Overlays to Achieve Data Persistence
During this presentation, we will present a P2P persistent data storage architecture. It exploits the properties of cluster-based peer-to-peer structured overlays together with a hybrid redundancy schema (a compound of light replication and rateless erasure coding) to guarantee durable access and integrity of data despite adversarial attacks. Prior to presenting this architecture, we will present the properties of rateless codes that make them appropriate for coping with communication channels having an unbounded loss rate. They are therefore very well suited to peer-to-peer systems.
This evaluation targets two goals. First, it compares the performance of both codes in different adversarial environments and in different application contexts. Second, it helps understanding how the parameters driving the behavior of the coding impact its complexity. We will then evaluate the triptych “availability – storage overhead – bandwidth usage” of our data storage architecture. We will show that despite massive attacks and high churn, it performs remarkably well.
Slides of the talk
Serge Defrance, Technicolor, Rennes
Home Networking as a Distributed File System View
Devices forming a Home Network have different capabilities and interfaces, discouraging users to organize their large digital content libraries. To help users, we propose to organize the Home Network according to a gateway-centric architecture, where the content access unification is realized at the file system level and where no additional software installation on devices is required. Solutions for realizing this unification individually exist for the various devices making up the Home Network (UPnP/DLNA devices, personal computers, cloud storage systems, etc). Unifying the content access at the file system level offers a powerful lever for many legacy applications, as far as these applications can access all shared data in the Home Network. Users can thus continue to use their PC’s file manager or favorite media player to browse or display shared content. An indexing application, running on the gateway, possibly managed by the ISP and accessible from any device via a simple web interface, enables more powerful content retrieval and user experience. Such application may be enriched to offer additional services like content format adaptation, duplication detection or automatic backup. Lastly we describe how this gateway-centric architecture can be leveraged by cloud applications such as distributed storage systems.
Slides of the talk
More information: Home Networking as a Distributed File System View, Serge Defrance, Remy Gendrot, Jean Le Roux, Gilles Straub and Thierry Tapie, SIGCOMM HomeNets 2011.
Remigiusz Modrzejewski, Project-Team MASCOTTE, I3S (CNRS/Université de Nice) / INRIA Sophia-Antipolis
Performance Analysis of P2P Storage Systems
Distributed or peer-to-peer storage systems have been proposed as an interesting solution to backup data in a scalable way at a low marginal cost. In this kind of systems redundancy is introduced to preserve the data in case of peer failures or departures. To ensure long-term fault tolerance, the storage system must have a self-repairing service that continuously reconstructs the fragments of redundancy that are lost. The speed of this reconstruction process is crucial for the data survival. This speed is mainly determined by how much bandwidth is available, which is a critical resource of such systems. In the literature, the durations of the reconstructions are modeled as independent (e.g., poissonian, deterministic, or more generally following any distribution). In practice, however, numerous reconstructions start at the same time (when the system detects that a peer has failed). Consequently, they are correlated to each other because concurrent reconstructions do compete for the same bandwidth. This correlation negatively impacts the efficiency of the bandwidth utilization and henceforth the repair time.
In this paper, we propose a new analytical framework that takes into account this correlation when estimating the repair time and the probability of data loss. Mainly, we introduce queuing models in which reconstructions are served by peers at a rate that depends on the available bandwidth. We show that the load is unbalanced among peers (young peers inherently store less data than the old ones). This leads us to introduce a correcting factor on the repair rate of the system. Lastly, we discuss scheduling and control issues. Indeed, each peer is involved in many reconstructions, so they need to schedule their order of processing. We compared different strategies and show
that a greedy-like policy gives smaller reconstruction times. The models and schemes proposed are validated by mathematical analysis, extensive set of simulations, and experimentation using the GRID’5000 test-bed platform.
Slides of the talk
More information: Analysis of Repair Time in Distributed Storage Systems, Frédéric Giroire, Sandeep Kumar Gupta, Remigiusz Modrzejewski, Julian Monteiro and Stéphane Perennes. INRIA Resarch Report RR-7538, 2011.