Availability and locality in distributed storage

Modern distributed file systems are deploying erasure codes to increase storage efficiency compared to block replication. Initially, classical error-correcting codes (like Reed-Solomon) were used but the benefits of custom code designs are now clear. Multiple companies including Microsoft and Facebook are designing and deploying distributed storage codes. Still, however, the fundamental information theoretic limits are not fully understood. Further, there has been significant activity on explicit code designs that are repair efficient and provide high data-availability. We present an overview of open problems and directions for modern distributed storage coding that offers higher data availability.