Source-linked AI summary

The Minimum Subpacketization of Reed-Solomon Codes with Linear Repair Achieving the Cut-Set Bound

Weijun Fang, Jing Qiu, Shu-Tao Xia, Fang-Wei Fu

arXiv:2610.37522v1cs.ITmath.IT

TL;DR

The paper studies the minimum subpacketization needed for Reed–Solomon codes to provide cut-set-optimal exact F_q-linear repair from any d helpers. It derives cross-ratio-based bounds and constructs codes from feasible degree assignments; in a specified parameter range, it determines the minimum exactly, while conjecturing the general formula.

  • Problem

    The minimum subpacketization for cut-set-optimal exact F_q-linear repair of every coordinate from every d-helper set is not determined by existing constructions.

    Source: §I · p. 2
  • Method

    The paper derives divisibility and relative-degree constraints on cross-ratio fields and constructs RS codes from feasible degree tuples.

    Source: B. Main Results · p. 3, §VI · p. 30
  • Results

    For every prime power q, ℓ_min(n,k,d;q)=sΔ_s(n,d)=sG_s(n) when d(d+2)≥4(n+1), including d=n−1.

    Source: B. Main Results · p. 3
  • Takeaways & Limitations

    The full-helper exact F_q-linear repair case resolves the corresponding case of Open Problem 20 posed by Ramkumar et al.

    Source: B. Main Results · p. 3

    The equality ℓ_min(n,k,d;q)=sG_s(n) does not hold for all admissible parameters; the paper conjectures the general formula ℓ_min(n,k,d;q)=sΔ_s(n,d).

    Source: §VI · p. 30

Abstract

from arXiv · show

In this paper, we study the minimum subpacketization of $(n,k)$ Reed-Solomon (RS) codes over $\mathbb{F}_{q^\ell}$ admitting cut-set-optimal exact $\mathbb{F}_q$-linear repair of every coordinate from any $d$ helpers. For $3\le k<d\le n-1$, write $s=d+1-k$ and denote this minimum by $\ell_{min}(n,k,d;q)$. Using divisibility and relative-degree constraints on cross-ratio fields of evaluation sets, we establish $\ell_{min}(n,k,d;q)\ge sG_s(d+1)$, where $G_s(m)$ is the minimum product of $m$ pairwise coprime integers, each at least $s$. Let $\Delta_s(n,d)$ be the minimum of $lcm(a_1,\ldots,a_n)$ over positive integers $a_1,\ldots,a_n$ satisfying $lcm(a_j:j\in J\cup\{i\})\ge s~lcm(a_j:j\in J)$ for every $d$-subset $J\subseteq[n]$ and every $i\in[n]\setminus J$. We construct RS codes with subpacketization $s\Delta_s(n,d)$ and further prove that $\ell_{min}(n,k,d;q)=s\Delta_s(n,d)=sG_s(n)$ whenever $d(d+2)\geq 4(n+1)$, for every prime power $q$. In particular, this determines the minimum subpacketization in the full-helper case $d=n-1$, resolving the corresponding exact linear-repair case of Open Problem 20 posed by Ramkumar et al. in their survey on codes for distributed storage. We further give a finite formula for $G_s(n)$ and determine the exact threshold for its closed-form expression in terms of the first $n$ primes. For fixed $k$ and $d$, we prove $\log\ell_{min}(n,k,d;q)= \Theta_{k,d}(\log n\log\log n)$ as $n\to\infty$, using divisor-function estimates and cover-free families. An affine-plane construction shows that $\Delta_s(n,d)<G_s(n)$ for some parameters, so $sG_s(n)$ is not a universal formula. We conjecture that $\ell_{min}(n,k,d;q)=s\Delta_s(n,d)$ for every prime power $q$ and all $3\le k<d\le n-1$.

I. INTRODUCTION

The paper asks how small the subpacketization of Reed–Solomon codes can be while preserving cut-set-optimal exact linear repair from every set of d helpers. It develops lower bounds and constructions to characterize this minimum, including an exact formula in a parameter range.

A. Reed-Solomon Codes and Linear Repair

  • Viewing each RS symbol over F_q^ℓ as a vector over F_q makes it an array code, but its scalar extension field and evaluation structure impose additional algebraic constraints.Cut-set-optimal repair downloads dℓ/s F-symbols in total.Source: §I · p. 2, A. Reed-Solomon Codes and Linear Repair · p. 5

I. INTRODUCTION

  • The study focuses on exact F-linear repair of every coordinate from every d-helper set, with bandwidth meeting the cut-set bound.The extension-field degree ℓ is the subpacketization being minimized.Source: §I · p. 2, B. Main Results · p. 3

A. Linear Repair of Reed-Solomon Codes

  • Earlier constructions attain cut-set-optimal repair, but do not determine the least subpacketization; this paper minimizes over evaluation sets and allows more general degree assignments.Prior work includes the Tamo–Ye–Barg construction, and a basis-transformation method for arbitrary distinct prime degrees.Source: A. Linear Repair of Reed-Solomon Codes · pp. 2–3

B. Main Results

  • The minimum subpacketization is exactly sΔ_s(n,d)=sG_s(n) for d(d+2)≥4(n+1), including the full-helper case d=n−1.This resolves the stated exact F_q-linear repair case of Open Problem 20.Source: B. Main Results · p. 3
  • For every admissible parameter set, RS codes achieve cut-set-optimal repair with subpacketization sΔ_s(n,d).The construction uses feasible degree tuples that may contain composite degrees and shared prime factors.Source: B. Main Results · p. 3
  • The paper gives a finite formula for G_s(m) and determines the threshold for its closed-form expression; for s=2, G_2(m)=p#_m.The threshold is expressed using the first m primes.Source: B. Main Results · p. 4
  • For fixed k and d, log ℓ_min(n,k,d;q)=Θ_{k,d}(log n log log n), while an affine-plane construction shows sG_s(n) is not universal.The paper conjectures that ℓ_min(n,k,d;q)=sΔ_s(n,d) for every prime power q and all 3≤k<d≤n−1.Source: B. Main Results · p. 4

C. Organization

  • The paper proceeds from repair preliminaries to cross-ratio lower bounds, degree-tuple constructions and arithmetic formulas, then fixed-parameter asymptotics; appendices include an affine-plane example.That example has Δ_s(n,d)<G_s(n).Source: C. Organization · p. 4

B. Finite-Field Degrees and Local Repair Subspaces

Finite-field composita let extension degrees be tracked through least common multiples. A basis transformation removes a congruence restriction on repair-subspace degree choices, while the lower-bound analysis combines cross-ratio degree constraints across helper sets.

B. Finite-Field Degrees and Local Repair Subspaces

  • Composita of finite fields have degree equal to the least common multiple of the component degrees.The paper applies this formula to relative extension degrees over fields generated by chosen helpers.Source: B. Finite-Field Degrees and Local… · pp. 5–6
  • The earlier construction required evaluation-point degrees congruent to 1 modulo s, restricting available choices.A basis-transformation argument in prior work removes this restriction and gives a subspace decomposition for every relative degree r ≥ s.Source: B. Finite-Field Degrees and Local… · p. 6

III. LOWER BOUNDS ON SUBPACKETIZATION

  • Cross-ratio degree constraints yield lower bounds sΔ_s(n,d) when 2d ≥ n+3 and sG_s(n) when d(d+2) ≥ 4(n+1).The stronger bound follows from a finer analysis of private and shared prime factors among the degrees.Source: §III · p. 6

A. Cross-Ratio Constraints

Cut-set-optimal full-helper repair imposes dimension and alignment conditions on evaluation spaces. These conditions identify a field generated by helper cross-ratios and force a relative-degree constraint for each failed coordinate.

A. Cross-Ratio Constraints

  • Full-helper repair implies s divides ℓ and yields an ℓ-dimensional polynomial subspace whose evaluation spaces support the repair.The failed coordinate's evaluation space is all of K, and the helper-space dimensions are ℓ/s.Source: A. Cross-Ratio Constraints · pp. 6–7
  • For any s helper points, every prescribed tuple from their evaluation spaces is attained by a polynomial in the repair subspace.This follows from injectivity of evaluation on polynomials of degree less than s and equality of the relevant dimensions.Source: A. Cross-Ratio Constraints · p. 7
  • Equality in the cut-set bound aligns helper evaluation spaces, and comparing the resulting identities yields cross-ratio relations among them.The aligned space is linear over the cross-ratio field of the helper points.Source: A. Cross-Ratio Constraints · pp. 8–9
  • The failed coordinate forces relative degree at least s over the helper cross-ratio field.The associated s elements are linearly independent over that field, yielding the relative-degree constraint used in later bounds.Source: A. Cross-Ratio Constraints · p. 10

B. A General Lower Bound

Cross-ratio fields impose divisibility and relative-degree constraints that give a full-helper subpacketization lower bound. Puncturing transfers the result to arbitrary helper sets, producing a general lower bound for RS repair.

B. A General Lower Bound

  • Deleting one evaluation point in a full-helper configuration costs relative degree at least s, motivating the arithmetic quantity G_s.G_s is defined through products of pairwise coprime integers constrained to be at least s.Source: B. A General Lower Bound · p. 11
  • For full-helper repair, there are n pairwise coprime integers at least s whose product gives the lower bound ℓ ≥ sG_s(n).The pairwise coprimality follows from compositum identities for the cross-ratio fields and the relative-degree constraints.Source: B. A General Lower Bound · pp. 11–12
  • For arbitrary d-helper repair, puncturing to d+1 coordinates gives the lower bound ℓ ≥ sG_s(d+1).Each coordinate in the punctured code has the other d coordinates as an allowed helper set in the original code.Source: B. A General Lower Bound · p. 12
  • For RS codes, the general lower bound is stronger than the cited scalar-MDS bound because the coprime factors each contribute at least one distinct prime divisor.The comparison is stated relative to Tamo, Ye, and Barg [18, Theorem 5].Source: B. A General Lower Bound · p. 12

C. An Improved Lower Bound for 2d ≥n + 3

When 2d ≥ n + 3, constraints from all failed-coordinate/helper-set pairs yield an LCM representation of cross-ratio-field degrees. This gives a lower bound on subpacketization in terms of the minimum feasible LCM, Δ_s(n,d).

C. An Improved Lower Bound for 2d ≥ n + 3

D. A Stronger Lower Bound for d(d + 2) ≥4(n + 1)

For d(d + 2) ≥ 4(n + 1), a finer analysis of cross-ratio degrees extends the lower bound to ℓ ≥ sG_s(n). The argument accounts for both private degree factors and shared prime factors.

D. A Stronger Lower Bound for d(d + 2) ≥ 4(n + 1)

  • For d(d + 2) ≥ 4(n + 1), the paper establishes the subpacketization lower bound ℓ ≥ sG_s(n).This extends the lower bound beyond the range 2d ≥ n + 3.Source: D. A Stronger Lower Bound for d(d + 2)… · pp. 15, 20
  • When overlaps are too small for the earlier LCM representation, the proof tracks how cross-ratio degrees change as evaluation points are deleted.It separates private factors associated with individual coordinates from prime factors shared across cross-ratio fields.Source: D. A Stronger Lower Bound for d(d + 2)… · p. 15
  • Nondeficient coordinates contribute pairwise-coprime factors g_i ≥ s, while deficient coordinates force additional distinct prime factors into the shared degree.For deficient coordinates, the repair constraint requires a missing factor from the shared part.Source: D. A Stronger Lower Bound for d(d + 2)… · pp. 15–17
  • The structural and arithmetic estimates combine to show that the cross-ratio degree is at least G_s(n) in the stated parameter range.Since that degree divides ℓ/s, the result yields the subpacketization lower bound.Source: D. A Stronger Lower Bound for d(d + 2)… · pp. 19–20

IV. CONSTRUCTION AND MINIMUM SUBPACKETIZATION

The section constructs RS codes with subpacketization sΔ_s(n,d), then combines the construction with lower bounds to identify the minimum in a specified parameter range.

IV. CONSTRUCTION AND MINIMUM SUBPACKETIZATION

  • The construction gives RS codes with subpacketization sΔ_s(n,d) for every admissible parameter choice.The construction is based on feasible degree tuples.Source: §IV · p. 20
  • Combining the construction with the lower bound determines the minimum subpacketization when d(d + 2) ≥ 4(n + 1).Source: §IV · p. 20
  • The section also derives an exact threshold for a closed-form expression for G_s(m) and a finite formula valid for all m.Source: §IV · p. 20

A. Construction from Feasible Degree Tuples

A feasible degree tuple determines evaluation points and an extension field for an RS code. The construction then gives exact F-linear repair at the cut-set bound for every required failure and helper-set choice.

A. Construction from Feasible Degree Tuples

  • Feasibility ensures that adjoining a failed evaluation point to the helper field has relative degree at least s, supplying the repair subspace required by the construction.The argument applies separately to each failed coordinate and helper set.Source: A. Construction from Feasible Degree… · p. 21
  • Given a feasible tuple, the construction sets M=lcm(a_1,...,a_n), uses E=F_(q^M), and chooses distinct evaluation points of degrees a_i.The resulting code has subpacketization ℓ=sM.Source: A. Construction from Feasible Degree… · p. 21
  • Dual parity checks supported on the failed coordinate and chosen helpers produce trace values that uniquely determine the missing symbol.The trace pairing is nondegenerate, so the values determine the failed coordinate.Source: A. Construction from Feasible Degree… · p. 22
  • Each helper sends F-linear functionals of its stored symbol, and the total repair bandwidth is dM=dℓ/s, exactly the cut-set bound.The construction provides this repair for every required coordinate and helper-set pair.Source: A. Construction from Feasible Degree… · pp. 21–22

B. Exact Minimum Subpacketization for d(d + 2) ≥4(n + 1)

The construction and lower bound coincide to give an exact minimum-subpacketization formula under the theorem’s condition. In the full-helper case, the theorem resolves the corresponding exact linear-repair question posed by Ramkumar et al.

B. Exact Minimum Subpacketization for d(d + 2) ≥4(n + 1)

C. Exact Formulas for Gs(m)

The section identifies when G_s(m) has a closed form and gives a finite minimization formula valid below that threshold as well. The threshold depends on the largest prime and prime power below s.

Closed-form threshold

  • The sufficient threshold m≥⌊s^2/4⌋ guarantees that the closed form applies.The section gives the exact threshold σ_s=π(u_s v_s), with ⌊s^2/4⌋ as a sufficient bound.Source: C. Exact Formulas for Gs(m) · pp. 23, 25
  • For m≥σ_s=π(u_s v_s), G_s(m) has the closed form obtained from the smallest prime powers at least s for primes below s, followed by successive primes at least s.Here u_s and v_s are the largest prime and largest prime power below s, respectively.Source: C. Exact Formulas for Gs(m) · pp. 23–24

Consequences and examples

Finite formula

  • For every m, the finite formula minimizes over disjoint nonempty sets of primes below s and uses the remaining factors as successive primes at least s.The allowed number of such sets ranges from zero to min{m,c_s}.Source: C. Exact Formulas for Gs(m) · pp. 26–27

Consequences and examples

  • For s=5 and m=2, combining the small primes 2 and 3 into 6 gives G_5(2)=6·5<5·7.This illustrates why factors using primes below s can improve on using only primes at least s.Source: C. Exact Formulas for Gs(m) · p. 27

V. ASYMPTOTICS FOR FIXED k AND d

For fixed k and d, the paper bounds the growth of minimum subpacketization using divisor-function estimates for a lower bound and cover-free families for an upper bound. Together, the bounds establish its asymptotic order, while leaving the optimal leading constant open.

Lower bound

  • The lower bound counts distinct subset least common multiples and uses divisor-function growth to bound their number.The proof converts the count into a lower bound on log ℓ_min.Source: §V · pp. 27, 28, 30

Asymptotic order

  • For fixed k and d, log ℓ_min(n,k,d;q)=Θ_{k,d}(log n log log n) as n grows.The theorem gives lower and upper bounds of this order.Source: §V · p. 29

Upper bound

  • The upper bound uses cover-free families together with the degree construction to obtain log ℓ_min(n,k,d;q)≤(C_d+o(1))log n log log n.The constant C_d depends only on d.Source: §V · p. 30

Open question

  • The optimal leading constant in the fixed-parameter asymptotic remains open.The order of growth is determined, but not its optimal leading constant.Source: §V · p. 30

VI. CONCLUSION

The paper studies minimum subpacketization for Reed-Solomon codes with cut-set-optimal exact linear repair, establishing its conjectured general formula in a specified parameter range. It also identifies settings where the simpler sG_s(n) expression is not universal.

VI. Conclusion

  • In the full-helper case, the results determine minimum subpacketization for exact linear repair and resolve the corresponding case of Open Problem 20.The case is d=n−1 with 3≤k<n−1.Source: §VI · p. 30
  • The paper also gives explicit formulas for G_s(n) and determines the asymptotic order of log ℓ_min(n,k,d;q) for fixed k and d..Source: §VI · p. 30
  • An affine-plane construction gives parameters with Δ_s(n,d)<G_s(n), so sG_s(n) is not a universal formula; the paper conjectures sΔ_s(n,d) for all admissible parameters.The conjecture is proved in the stated quadratic range.Source: §VI · p. 30
  • The paper proves its conjectured formula for minimum subpacketization whenever d(d+2)≥4(n+1).Outside this range, the matching lower bound remains to be established.Source: §VI · p. 30

APPENDIX A PROOF OF LEMMA 10

Appendix A proves Lemma 10 using bounds on binomial coefficients and prime counts. Appendix B uses an affine-plane construction to show that Δ_s(n, d) can be strictly smaller than G_s(n), so the corresponding equality is not universal.

APPENDIX A PROOF OF LEMMA 10

  • The proof bounds the central binomial coefficient using its size relative to the average binomial coefficient and an upper bound involving π(x).It uses the factorial valuation formula to bound the prime valuations of the central binomial coefficient.Source: Appendix A · p. 31
  • For t ≥ 4, the proof obtains π(x_t) > 4t^2 by applying a prime-counting bound to x_t = (t + 2)!.It establishes the needed inequality by induction on t.Source: Appendix A · p. 31

APPENDIX B STRICT INEQUALITY BETWEEN ∆s(n, d) AND Gs(n)

  • An affine-plane construction gives feasible degree tuples with Δ_s(n, d) < G_s(n) for some parameters, ruling out ℓ_min(n, k, d; q) = sG_s(n) as a universal formula.The passage states this conclusion together with Theorem 6.Source: Appendix B · p. 32
  • The construction selects nonvertical lines, some vertical lines, and singleton sets to form n members whose union has size max{T^2, n − T}.Pairwise-coprime weights make the least common multiple of any subfamily equal to the product of weights on its union; the resulting tuple is feasible.Source: Appendix B · p. 32
Loading 2610.37522v1…