Skip to Main content Skip to Navigation

Conception d'un système de partage de données adapté à un environnement de Fog Computing

Bastien Confais 1, 2, 3
1 RIO - Réseaux pour l'Internet des Objets
LS2N - Laboratoire des Sciences du Numérique de Nantes
3 STACK - Software Stack for Massively Geo-Distributed Infrastructures
Inria Rennes – Bretagne Atlantique , LS2N - Laboratoire des Sciences du Numérique de Nantes
Abstract : Utility Computing has evolved for many years leading to the infrastructure we know today as Cloud Computing. Nevertheless, these infrastructures are unable to satisfy the needs of the Internet of Things which requires low latency computing despite limited resources. In 2012, Cisco proposed a paradigm called Fog Computing, consisting of deploying a huge number of small servers, spread on many sites located at the edge of the network, close to the end devices. In this thesis, we try to create a seamless storage solution between the different Fog sites. Our first contribution consists in comparing existing storage solution and check if they can be used in a such environment. We show that InterPlanetary FileSystem (IPFS), an object store relying on a BitTorrent like protocol and a Distributed Hash Table is a promising solution. Nevertheless, the amount of network traffic exchanged between the sites to locate the data is important and has a non-negligible impact on the overall performance. Our second contribution consists in coupling IPFS with RozoFS, a distributed filesystem deployed on each site to limit the use of the DHT when accessed data are stored on the local site. Finally, we proposed to replace the distributed hash table by a location mechanism relying on a shortest path tree built on the physical topology, in order to contain the network traffic and to first request nodes at a close location, reachable with a low latency. By performing many experiments on the Grid’5000 testbed, we show that the coupling of IPFS with a Scale-Out NAS reduces by 34 % in average the access times and that our protocol to locate the objects reduces by 20 % the time to locate the data.
Complete list of metadata

Cited literature [429 references]  Display  Hide  Download
Contributor : Bastien Confais Connect in order to contact the contributor
Submitted on : Wednesday, September 5, 2018 - 12:34:00 PM
Last modification on : Wednesday, April 27, 2022 - 4:23:55 AM
Long-term archiving on: : Thursday, December 6, 2018 - 2:56:14 PM


  • HAL Id : tel-01868308, version 1


Bastien Confais. Conception d'un système de partage de données adapté à un environnement de Fog Computing. Calcul parallèle, distribué et partagé [cs.DC]. Université de Nantes, 2018. Français. ⟨tel-01868308⟩



Record views


Files downloads