This project studies fundamental problems in data coding that can improve the efficiency of distributed storage systems by increasing data reliability and availability while reducing storage overhead compared to existing industry standards. The results of this research can benefit storage applications ranging from financial, scientific monitoring, and signal processing to social networks and sharing platforms. The new combinatorial, coding, and information theoretic tools developed in this project will be incorporated in course curricula in the respective institutions of the principal investigators.
Data coding with locality, the focus of this project, is a rapidly developing area of coding theory that was initially motivated by applications in distributed storage, and has links to many areas of network science (e.g., index coding and network coding) as well as to computer science. This project advances the theory and practice of data coding with local recovery by investigating broad implications of the locality constraint in coding problems. These include studying new error-correcting code families and their decoding, fundamental limitations on the code parameters and capacity bounds under the requirements of local data recovery. The newly designed coding schemes developed in this project will be validated through implementation and evaluation in simulated computer environment, aiming at enhanced performance of data coding in current industry solutions.
This is a three-year, $250K grant.