Implementation and Performance Evaluation of Distributed Cloud Storage Solutions using Random Linear Network Coding

Research output: Research - peer-reviewArticle in proceeding

Abstract

This paper advocates the use of random linear network coding for storage in distributed clouds in order to reduce storage and traffic costs in dynamic settings, i.e. when adding and removing numerous storage devices/clouds on-the-fly and when the number of reachable clouds is limited. We introduce various network coding approaches that trade-off reliability, storage and traffic costs, and system complexity relying on probabilistic recoding for cloud regeneration. We compare these approaches with other approaches based on data replication and Reed-Solomon codes. A simulator has been developed to carry out a thorough performance evaluation of the various approaches when relying on different system settings, e.g., finite fields, and network/storage conditions, e.g., storage space used per cloud, limited network use, and limited recoding capabilities. In contrast to standard coding approaches, our techniques do not require us to retrieve the full original information in order to store meaningful information. Our numerical results show a high resilience over a large number of regeneration cycles compared to other approaches.
Close

Details

This paper advocates the use of random linear network coding for storage in distributed clouds in order to reduce storage and traffic costs in dynamic settings, i.e. when adding and removing numerous storage devices/clouds on-the-fly and when the number of reachable clouds is limited. We introduce various network coding approaches that trade-off reliability, storage and traffic costs, and system complexity relying on probabilistic recoding for cloud regeneration. We compare these approaches with other approaches based on data replication and Reed-Solomon codes. A simulator has been developed to carry out a thorough performance evaluation of the various approaches when relying on different system settings, e.g., finite fields, and network/storage conditions, e.g., storage space used per cloud, limited network use, and limited recoding capabilities. In contrast to standard coding approaches, our techniques do not require us to retrieve the full original information in order to store meaningful information. Our numerical results show a high resilience over a large number of regeneration cycles compared to other approaches.
Original languageEnglish
Title of host publicationCommunications Workshops (ICC), 2014 IEEE International Conference on
PublisherIEEE Press
Publication date2014
Pages249-254
ISBN (Print)9781479946402
DOI
StatePublished - 2014
Publication categoryResearch
Peer-reviewedYes
Event2014 IEEE International Conference on Communications - Sydney , Australia
Duration: 10 Jun 201414 Jun 2014
Conference number: 31675

Conference

Conference2014 IEEE International Conference on Communications
Nummer31675
LandAustralia
BySydney
Periode10/06/201414/06/2014
SeriesIEEE International Conference on Communications
ISSN1938-1883

Projects

ID: 209515787