Distributed storage based on sparse-graph codes
The recent years have witnessed the explosion of services and applications such as email, cloud computing, social and media networks, and video sharing. The grand challenge is to store, process and transfer this massive amount of data. According to recent estimates, the data to be stored grows at a rate of 45% per year, a factor of 40 by 2024. Data centers typically employ a collection of cheap devices/nodes, which are connected forming a so-called distributed storage (DS) system, where a central server keeps track of where each file is stored and performs all network operations. Every day, knowingly or unknowingly people connect to various private and public DS systems. It is nowadays accepted that current DS systems cannot face the unabated growing of digital data volume. Nextgeneration storage systems desperately need new technologies to improve their robustness to node failures, storage efficiency, complexity, and cost efficiency, in order to sustain the information revolution of modern societies. This project addresses the challenges above using modern coding theory and optimization theory. Our focus is on the application of sparse-graph erasure correcting codes to large DS networks. The proposed project falls into the category of fundamental research, but it is driven by practical concerns, and we expect our findings to give new and relevant insights for the design of practical coding schemes for next-generation DS systems.
Alexandre Graell i Amat (contact)
Professor at Chalmers, Electrical Engineering, Communication and Antenna Systems, Communication Systems
European Commission (EC)
Funding Chalmers participation during 2015–2017