[Paper Review] Optimal Fractional Repetition Codes

Natalia Silberstein, Tuvi Etzion|arXiv (Cornell University)|Jan 19, 2014

Advanced Data Storage Technologies31 references10 citations

TL;DR

This paper proposes optimal fractional repetition (FR) codes for distributed storage systems that enable uncoded, minimum-bandwidth repair via table-based node recovery. By leveraging combinatorial designs and regular/biregular graphs, the authors establish bounds on FR capacity and construct codes that achieve these bounds, while also enabling parallel data reads through connections to combinatorial batch codes and defining a rate hierarchy analogous to generalized Hamming weights.

ABSTRACT

Fractional repetition (FR) codes is a family of codes for distributed storage systems that allow for uncoded repair having the minimum repair bandwidth. However, in contrast to minimum bandwidth regenerating codes, where a random set of certain size of available nodes is used for a node repair, the repairs with FR codes are table based. In this work we consider bounds on the fractional repetition capacity. Optimal FR codes which attain these bounds are presented. The constructions of optimal FR codes are based on combinatorial designs and on different families of regular and biregular graphs. Finding optimal codes raises some interesting questions in graph theory. We discuss these questions and their solutions. In addition, we analyze other properties of the constructed codes, allowing parallel independent reads of many subsets of the stored symbols, by showing a connection to combinatorial batch codes. We also define for each code a rate hierarchy which resembles to the well known generalized Hamming weight hierarchy.

Motivation & Objective

• To establish theoretical bounds on the fractional repetition capacity of distributed storage codes.
• To construct optimal FR codes that achieve these bounds using combinatorial designs and regular/biregular graphs.
• To enable parallel, independent data reads by linking FR codes to combinatorial batch codes.
• To define a rate hierarchy for FR codes resembling the generalized Hamming weight hierarchy.

Proposed method

• Constructing FR codes using combinatorial designs such as balanced incomplete block designs (BIBD) and pairwise balanced designs.
• Utilizing regular and biregular graphs to model node replication and ensure uniform repair properties.
• Defining a table-based repair mechanism where specific node sets are precomputed for repair, ensuring minimal bandwidth.
• Mapping code structures to graph-theoretic objects to analyze and optimize code parameters.
• Establishing a rate hierarchy by analyzing the generalized Hamming weights of the code's dual code.
• Analyzing parallel read capabilities by relating the code structure to combinatorial batch code properties.

Experimental results

Research questions

• RQ1 What are the theoretical upper bounds on the fractional repetition capacity of distributed storage codes?
• RQ2 How can optimal FR codes be constructed to achieve these bounds using combinatorial and graph-theoretic methods?
• RQ3 What is the relationship between FR codes and combinatorial batch codes in enabling parallel independent data reads?
• RQ4 How can a rate hierarchy for FR codes be defined, and how does it relate to the generalized Hamming weight hierarchy?
• RQ5 What new graph-theoretic problems arise in constructing optimal FR codes, and how are they resolved?

Key findings

• Optimal FR codes achieving the theoretical capacity bounds are constructed using combinatorial designs and regular/biregular graphs.
• The constructions ensure uncoded repair with minimum repair bandwidth by predefining repair tables based on code structure.
• Parallel independent reads of multiple data subsets are supported through a direct connection to combinatorial batch code properties.
• A rate hierarchy analogous to the generalized Hamming weight hierarchy is defined for FR codes, enabling analysis of code performance across different data access patterns.
• The problem of constructing optimal FR codes leads to novel questions in graph theory, such as the existence of specific biregular graphs with desired properties, which are resolved in the paper.