Optimal repairing schemes for Reed-Solomon codes with alphabet sizes linear in lengths under the rack-aware model

Lingfei Jin,Gaojun Luo,Chaoping Xing

DOI: https://doi.org/10.48550/arXiv.1911.08016

2019-11-19

Abstract:In modern practical data centers, storage nodes are usually organized into equally sized groups, which is called racks. The cost of cross-rack communication is much more expensive compared with the intra-rack communication cost. The codes for this system are called rack-aware regenerating codes. Similar to standard minimum storage regenerating (MSR) codes, it is a challenging task to construct minimum storage rack-aware regenerating (MSRR) codes achieving the cut-set bound. The known constructions of MSRR codes achieving the cut-set bound give codes with alphabet size $q$ exponential in the code length $n$, more precisely, $q=\Omega(\exp(n^n))$. The main contribution of this paper is to provide explicit construction of MSRR codes achieving the cut-set bound with the alphabet size linear in $n$. To achieve this goal, we first present a general framework to repair Reed-Solomon codes. It turns out that the known repairing schemes of Reed-Solomon codes can be realized under our general framework. Several techniques are used in this paper. In particular, we use the degree decent method to repair failure node. This technique allows us to get Reed-Solomon codes with the alphabet size linear in $n$. The other techniques include choice of good polynomials. Note that good polynomials are used for construction of locally repairable code in literature. To the best of our knowledge, it is the first time in this paper to make use of good polynomials for constructions of regenerating codes.

Information Theory

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in modern data centers, storage nodes are usually organized into groups of equal size (called racks), and the cost of cross - rack communication is much higher than that of intra - rack communication. In this case, how to construct the Minimum - Storage Rack - aware Regenerating (MSRR) codes to achieve the optimal repair bandwidth. Specifically, the goal of the paper is to provide an explicit method for constructing MSRR codes so that these codes can reach the cut - set bound, while the size of the alphabet is linearly related to the code length. Before this, the known construction methods of MSRR codes usually give an alphabet size that is exponential, which is not ideal in practical applications.

The main contributions of the paper are as follows:
1. A general framework for repairing Reed - Solomon codes is proposed. In particular, in the rack - aware model, a faulty node can be repaired with the help of any other \(d\) racks.
2. Three types of Reed - Solomon codes are designed based on good polynomials, and the repair scheme is implemented in the rack - aware model. These codes reach the rack - aware cut - set bound, and the size of the alphabet is linearly related to the code length.
3. By using the polynomial degree - reduction method, the problem of excessive polynomial degree caused by using good polynomials is solved, so that the size of the alphabet can be reduced while maintaining the codeword length.

Through these techniques, the paper successfully reduces the alphabet size of MSRR codes from the previous exponential level to the linear level, which is of great significance for improving the efficiency and reducing the cost of data storage systems.