Grant: Russian Foundation for Basic Research
Grant period: 2018–2019
Project instructor: A. Frolov
Description: Distributed storage systems have reached such a massive scale that recovery from several disk failures is now a regular operation of the system, rather than an exception. To ensure reliable storage, error correcting codes are used. Note, that the most common scenario is a single disk failure, therefore constructing codes (in what follows Locally Recoverable Codes, LRC) that ‘optimize’ the recovery of a single disk becomes an important problem. The main requirement is to minimize the total number of disks accessed during the repair process. This topic immediately became popular due to obvious applications in distributed storage systems. We will consider LRC generalizations that ensure load balancing in distributed storage systems. We already derived bounds on the parameters of such codes, our main goal is to propose explicit constructions, that can be used in industry.