On Maximally Recoverable Local Reconstruction Codes

In recent years the explosion in the volumes of data being stored online has resulted in distributed storage systems transitioning to erasure coding based schemes. Local Reconstruction Codes (LRCs) have emerged as the codes of choice for these applications. An (n, r ,h,a,q)-LRC is a q-ary code, where encoding is as a two stage process. In the first stage, h redundant parity symbols are generated from k data symbols. In the second stage, the k+h symbols are partitioned into sets of size r −a and each set is extended with a redundant symbols using an MDS code to form a local group. Local groups ensure that when at most a coordinates are erased, any missing coordinate can be recovered by accessing at most r − a symbols. Also, if a larger number of coordinates is erased, the missing symbols can be recovered by potentially accessing all remaining symbols. An (n, r ,h,a,q)-LRC code as above is Maximally Recoverable (MR), if it corrects all erasure paŠerns which are information theoretically correctable given the presence of local groups. Obtaining MR LRCs over finite fields ofminimal size is important in practice and has been the goal of a line of work in coding theory. In this work we make progress towards this goal. In particular: – We show that when a and h are constant and r may grow, for every maximally recoverable LRC, q > Ωa,h ( n · rmin{a,h−2} ) . Prior to our work, there was no super-linear lower bound known on the field size of MR LRCs for any seŠing of parameters. – We obtain a family of MR (n, r ,h = 2,a,q)-LRCs, where q = O(n) for all seŠings of parameters. Prior to our work the best constructions required q to be quadratic in n for some regimes. – We obtain a family of MR (n, r ,h = 3,a,q)-LRCs, where q = O(n) for all seŠings of parameters. Prior to our work the best constructions required q to be nΘ(a) for some regimes. – Our results in the first two bullets above suggest the seŠing of r = 3,a = 1,h = 3 as the first seŠing where existence of MR LRCs over fields of near linear size is an open question. We resolve this question in the positive by developing a new approach to LRC constructions based on elliptic curves and arithmetic progression free sets. ∗Research supported by NSF CAREER award 1451191 and NSF grant CCF-1523816. Most of this work was done when the author was visiting Microso‰ Research. †Research supported in part by NSF grant CCF-1563742. Most of this work was done during a visit by the author to Microso‰ Research, Redmond. Œe work was also partly done when the author was visiting the School of Physical and Mathematical Sciences, Nanyang Technological University, Singapore.