Storage

Exact-MSR Codes for Distributed Storage With Low Repair Complexity

Date Added: Mar 2012
Format: PDF

Distributed storage nodes connected in a network can improve the system reliability against node failures, when redundancy is introduced to ensure data recovery. Two types of data recovery are of interest: one is the data recovery of the entire message file, called data reconstruction, and the other is how to repair partial messages stored in some nodes, referred to as data repair. It was shown that there exists a tradeoff between repair bandwidth (the number of packets downloaded from other nodes for data recovery) and storage needed at nodes, and codes that achieve the tradeoff curve are called regenerating codes.