Peer-to-peer; Cloud Computing; Overlay Networks; Distributed Storage
Schiavoni Valerio, Felber Pascal, Rivière Etienne (2012), [Poster] User-Space Network Emulation for the SPLAY Platform, in Poster session, ASPLOS 2012
, ACM, USA.
Charles Lucas, Felber Pascal, Halalai Raluca, Rivière Etienne, Schiavoni Valerio, Valerio José (2012), [Technical report] An overview of new features in the SPLAY framework for simple distributed systems evaluation
, Université de Neuchâtel, Neuchâtel.
Matos Miguel, Schiavoni Valerio, Felber Pascal, Oliveira Rui, Riviere Etienne (2012), BRISA: Combining Efficiency and Reliability in Epidemic Data Dissemination, in 26th IEEE International Parallel & Distributed Processing Symposium (IPDPS)
, 26th IEEE International Parallel & Distributed Processing Symposium (IPDPS), Shangai, China.
Valerio José, Felber Pascal, Rajman Martin, Rivière Etienne (2012), CADA: Collaborative Auditing for Distributed Aggregation, in 9th European Dependable Computing Conference (EDCC)
, Sibiu, RomaniaIEEE, USA.
Felber Pascal, Kropf Peter, Leonini Lorenzo, Luu Toan, Rajman Martin, Rivière Etienne, Schiavoni Valerio, Valerio José (2011), CoFeed: privacy-preserving Web search recommendation based on collaborative aggregation of interest feedback, in Software Practice and Experience
Schiavoni Valerio, Rivière Etienne (2011), WHISPER : communications confidentielles dans un monde décentralisé, in AlgoTel 2011
, Cap Estérel, FranceHAL, France.
Schiavoni Valerio, Rivière Etienne, Felber Pascal (2011), Whisper: Middleware for Confidential Communication in Large-Scale Networks, in 31st International Conference on Distributed Computing Systems
, Minneapolis, MN, USA31th IEEE International Conference on Distributed Computing Systems (ICDCS), Minneapolis, Minnesota, USA.
Schiavoni Valerio (2010), [Poster] The Private Peer Sampling Service -- The Ground for your Secret Society, in Poster session, OSDI 2010
, Vancouver, BC, CanadaUSENIX, USA.
Felber Pascal, Rajman Martin, Rivière Etienne, Schiavoni Valerio, Valerio José (2010), SPADS: Publisher Anonymization for DHT Storage, in 10th IEEE International Conference on Peer-to-Peer Computing (P2P'10)
, Delft, NetherlandsIEEE, USA.
Matos Miguel, Schiavoni Valerio, Felber Pascal, Oliveira Rui, Rivière Etienne, Lightweight, Efficient, Robust Epidemic Dissemination, in Journal of Parallel and Distributed Computing (JPDC)
Computing is becoming increasingly pervasive, parallel, and distributed. We are witnessing a significant evolution in the scale and nature of computing systems, with great amounts of resources that currently remain largely unharnessed, uncontrolled, and isolated. Already, research and industry are starting to approach this new reality, notably by the ``cloud computing'' initiative that, despite being surrounded by much hype, investigates new problems and lays down conceptual foundations for federating the vast populations of computing, storage, data, and software resources. The general goal is to harness their power and to control their complexity: exploiting parallelism, distribution and heterogeneity, facilitating collaborative and self organizing behavior, easing resource sharing, and ensuring security. For this an end-to-end, system-wide approach is necessary.As part of a just completed SNF project, we have built a generic deployment infrastructure named SPLAY that forms a sound basis to building a lightweight and efficient cloud platform: it provides lightweight server virtualization, with a sandboxed and OS-agnostic execution environment. The goal of this project is to design and implement the next generation of SPLAY, with novel extensions, enhancements, and services dedicated to supporting cloud applications.Providing a complete general-purpose cloud computing platform is far beyond our objectives. Instead, we are interested in developing a simple yet easily extensible peer-to-peer infrastructure that allows for designing emerging applications and services that can take advantage of the cloud computing paradigm. The planned MistNet platform should be understood as an infrastructure and associated services for a lightweight experimental cloud.We plan to address the following three challenges in this project: First, we will provide high-level facilities for designing distributed data structures, and mapping these structures over the physical networked cloud, hence forming overlay networks. Specifically, we plan to use a domain-specific language, to express not only structured overlays, but also self-structured peer-to-peer networks that are most relevant for cloud computing. Noteworthy, our approach will distinguish the management of the data structure and its mapping to physical nodes as two separate operations, allowing more flexibility and greater abstraction.Second, we will investigate new mechanisms for fully decentralized and dynamically scalable storage facilities for applications operating on the cloud. Our main objectives are low latency, high availability, privacy and security. The placement and distribution of the data will be driven by constraints on quality of service as well as the type of access patterns to be supported. We will incorporate efficient querying and indexing facilities. Specifically, we will investigate models that can combine classical immediate queries and continuous queries (which are instrumental for running ``bag-of-tasks'' applications in the cloud).Finally, we will evolve the SPLAY infrastructure to support cloud-oriented applications by providing the basic services of a decentralized operating system, with support for code execution and resource management. We will put particular emphasis on security, e.g., to prevent tampering with data and to allow for verification of computation results. The final objective is to have a secure virtualized execution environment deployable pervasively across different ranges of devices and that collectively participate to an experimental research cloud.To sum up, the MistNet experimental cloud will consist of a service-oriented peer-to-peer architecture with dedicated support for communication and overlay management, data storage, and code execution.