On binary locally repairable codes with distance four
Introduction
Locally repairable codes are a family of erasure codes that can recover any code symbol by accessing other survived code symbols. Such codes can be used in distributed storage systems to improve repair efficiency. The concept of codes with locality was introduced by Gopalan et al. [1], Oggier and Datta [2], and Papailiopoulos et al. [3], [4]. For a q-ary linear code , if for any ∈ , the i-th code symbol can be recovered by accessing no more than other code symbols, is said to have locality . Further, has locality r if all its symbols have locality at most r, and it is denoted as . If , we write it as .
In [1], a Singleton-like bound of an was proved as: If , the bound (1) degenerates to the classical Singleton bound. Like the classical Singleton bound, the Singleton-like bound (1) does not take into account the cardinality of the code alphabet size q, and it is not tight in many cases. A bound taking field size into consideration was presented in [5], which is called Cadambe-Mazumdar (CM) bound. This bound says that an code satisfies where is the largest possible dimension of a code of length n, for given field size q and minimum distance d. If , will be written as .
Many constructions of LRCs achieving the upper bound of (1) or (2) have been proposed in [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31] and references therein. Among these work, binary LRCs (BLRCs) receive much more attention, since they are easily implemented and no multiplications are needed in encoding, decoding and repair, see [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21].
BLRCs with that meet the CM bound (2) have been constructed by using anti-codes [6]. Authors of [7], [8], [9], [10] proposed constructions of BLRCs with specific parameters by using cyclic codes. Authors of [11], [12], [13], [14], [15], [16], [17] discussed constructions of BLRCs from classical codes and d-optimal codes and obtained many codes meeting the CM bound (2).
In [18], the authors focused on BLRCs with and locality . Those BLRCs achieved the bound (1) when and achieved the bound (2) when , all of which satisfy . They provided a mean-time to data-loss (MTTDL) analysis on their codes and showed that was sufficient such that the reliability of LRCs surpassed that of the widely used 3-replication method for most practical scenarios. In [15], Rao determined the locality for binary codes with for and showed that all these BLRCs achieved the bound (2).
Hao et al. gave constructions of LRCs from parity check matrices of linear codes in [19] and [20]. In [20], BLRCs achieving the bound (1) were discussed and a class of BLRCs with d=4 and achieving the bound (2) was also constructed. The authors of [20] claimed that there were only four classes of BLRCs with distance 2 or 4 attaining the bound (1), yet this claim is incomplete. In [21], results of [19] and [20] were further improved, a missing class of BLRCs attaining the bound (1) in [20] was added there. Here, we list BLRCs attaining the bound (1) in [12] and [21], and known BLRCs with distance four in [20], [18] and [15].
Proposition 1 [21], [18], [12] There are only five classes of LRCs with attaining the Singleton-like bound: (1) for and ; (2) for and ; (3) for ; (4) for ; (5) , , and .
Proposition 2 There are the following BLRCs with distance four. (1) ([20]) If and , then there are BLRCs attaining the CM bound. (2) ([15]) If and , then there are BLRCs for and , these BLRCs achieve the CM bound.
An code is d-optimal if there is no code. The BLRCs of Proposition 2 (2) are all d-optimal codes. According to [32], the LRCs given in Proposition 1 only include 14 d-optimal codes, namely LRCs for , LRCs for and LRC. The others are not d-optimal codes. There are some d-optimal codes in Proposition 2 (1) when r and l are chosen properly.
To determine the optimality of LRCs, we give the following definition:
Definition An LRC is optimal if it achieves one of the following conditions: (1) attains the Singleton-like bound. (2) attains the CM bound. (3) An LRC does not exist.
Inspired by those constructions in [18], [20] and [15], we will focus on constructing BLRCs with distance four and determine their optimality.
The contribution to this paper is three-fold. Firstly, we determine code length with which an code exists, then construct LRC with r as small as possible. We also show that our LRCs with are all optimal BLRCs. Secondly, we give criterion for which LRCs attain the CM bound with , and prove that most of the obtained BLRCs can achieve the CM bound. Finally, for , we determine all d-optimal codes that can achieve the CM bound.
The rest of this paper is organized as follows. In the next section, we begin with some preliminaries on linear codes and LRCs, and known results on judging the locality of linear codes. In Section 3, we present our main results on construction of BLRCs, and determine their optimality. Section 4 makes comparison with related results. The last section is the conclusion.
Section snippets
Preliminaries
This section introduces basic concepts on linear codes and some results on locally repairable codes [33], [34], [1], [25]. At first, we give some notations for later use.
(i) Let , for and be called as interval. Denote and as the all-one and all-zero row vectors, whose transpose are denoted as and , respectively.
(ii) Denote , , , and . It is obvious that the columns of are the
Construction of LRCs and their optimality
In this section, we will give construction of code for given parameters and , denote and call m the check-dimension of . We always assume that is a code without all zero coordinates, since a code with all zero coordinates, that is to say, there exists a coordinate which is not covered, cannot be used for local repair.
If , there is only one code, this code has locality 1. If , according to [12], there are three codes without all zero
A comparison with known results
In this section, we will compare our LRCs with those in [15], [20], [18], [9] and [11]. BLRCs with parameters obtained in these references are listed in Table 1.
We will show that the following claim holds:
(I) The BLRCs with No.1-No.6 are special cases of our construction.
(II) Some BLRCs with No.7-No.14 can be improved by our construction.
Conclusions and discussions
In this paper, we have investigated construction of binary LRCs with good locality for all lengths and dimensions. Our constructions are more flexible and can offer wider choices of the code parameters, i.e., code length, dimension, and locality. Although almost all of our codes are optimal BLRCs, there are some codes whose optimality can not be determined. In the future, we will study how to improve the locality of the BLRCs that cannot attain the CM bound or prove their optimality by
CRediT authorship contribution statement
Ruihu Li: Conceptualization, Funding acquisition, Methodology, Writing – original draft, Writing – review & editing. Sen Yang: Software, Validation, Writing – review & editing. Yi Rao: Supervision, Writing – review & editing. Qiang Fu: Funding acquisition, Writing – review & editing.
Acknowledgement
The authors are very grateful to the reviewer and the FFA's Editor, for their detailed comments and suggestions that much improved the presentation and quality of this paper. This work is supported by National Natural Science Foundation of China under Grant No. 11801564 and 11901579.
References (34)
- et al.
Locality of optimal binary codes
Finite Fields Appl.
(2017) - et al.
On the locality of codeword symbols
IEEE Trans. Inf. Theory
(2012) - et al.
Self-repairing homomorphic codes for distributed storage systems
- et al.
Simple regenerating codes: network coding for cloud storage
- et al.
Locally repairable codes
IEEE Trans. Inf. Theory
(2014) - et al.
Bounds on the size of locally recoverable codes
IEEE Trans. Inf. Theory
(2015) - et al.
Optimal binary locally repairable codes via anticodes
- et al.
Binary cyclic codes that are locally repairable
- et al.
Optimal linear and cyclic locally repairable codes over small fields
- et al.
Cyclic LRC codes and their subfield subcodes
Cyclic linear binary locally repairable codes
Binary linear locally repairable codes
IEEE Trans. Inf. Theory
Bounds and constructions for linear locally repairable codes over binary fields
Optimal binary linear locally repairable codes with disjoint repair groups
SIAM J. Discrete Math.
Codes for big data storage and application
Binary locally repairable codes with minimum distance at least 6 based on partial t-spreads
IEEE Commun. Lett.
New constructions of binary and ternary locally repairable codes using cyclic codes
IEEE Commun. Lett.
Cited by (4)
A characterization of optimal locally repairable codes
2023, Discrete MathematicsOn locality of binary distance-optimal codes
2024, Cryptography and CommunicationsOn locality of binary distance-optimal codes
2022, Research Square