Iterative rounding approximation algorithms for degree-bounded node-connectivity network design

We consider the problem of finding a minimum edge cost subgraph of a graph satisfying both given node-connectivity requirements and degree upper bounds on nodes. We present an iterative rounding algorithm of the biset LP relaxation for this problem. For directed graphs and $k$-out-connectivity requirements from a root, our algorithm computes a solution that is a 2-approximation on the cost, and the degree of each node $v$ in the solution is at most $2b(v) + O(k)$ where $b(v)$ is the degree upper bound on $v$. For undirected graphs and element-connectivity requirements with maximum connectivity requirement $k$, our algorithm computes a solution that is a $4$-approximation on the cost, and the degree of each node $v$ in the solution is at most $4b(v)+O(k)$. These ratios improve the previous $O(\log k)$-approximation on the cost and $O(2^k b(v))$ approximation on the degrees. Our algorithms can be used to improve approximation ratios for other node-connectivity problems such as undirected $k$-out-connectivity, directed and undirected $k$-connectivity, and undirected rooted $k$-connectivity and subset $k$-connectivity.

💡 Research Summary

The paper tackles a fundamental network design problem that simultaneously imposes node‑connectivity requirements and upper bounds on the degree of each node. Formally, given a graph G = (V, E) with edge costs, a function b : V → ℕ specifying a degree limit for each vertex, and connectivity demands that require at least k internally vertex‑disjoint paths between prescribed pairs of vertices (or from a root to all vertices), the goal is to find a minimum‑cost subgraph H ⊆ G that satisfies both sets of constraints. This problem models many practical scenarios such as resilient communication networks, power‑grid restoration, and data‑center interconnects, where both fault tolerance (connectivity) and equipment limits (degree) matter.

Prior work typically treated these constraints separately. For pure node‑connectivity, approximation algorithms achieve O(log k) cost ratios using cut‑based linear programs, but they ignore degree limits. When degree bounds are added, existing methods either inflate the cost by a logarithmic factor or cause the degree violation to blow up exponentially (O(2^k b(v))). Consequently, there was a clear gap: no algorithm simultaneously offered a constant‑factor cost guarantee and a modest, preferably linear, violation of the degree bounds.

Technical contribution: The authors introduce a biset linear programming (LP) relaxation that captures node‑connectivity more precisely than ordinary cut‑LPs. A biset is an ordered pair (S, T) where S ⊆ T ⊆ V; edges crossing from S to V\T represent the “cut” of the biset. The biset LP enforces that for every biset (S, T) relevant to the connectivity requirement, the sum of selected edge variables crossing the biset is at least the required connectivity value. Degree constraints are added as simple linear inequalities ∑_{e∈δ(v)} x_e ≤ b(v). This formulation yields a fractional solution x* that respects both connectivity and degree limits in a relaxed sense.

The core algorithm is an iterative rounding scheme applied to the biset LP. At each iteration the LP is solved to optimality, and the following actions are taken:

Large variable fixing – any edge e with x*_e ≥ ½ is rounded up to 1 and permanently added to the solution.
Degree‑tight vertex handling – if a vertex v already exceeds its degree bound in the partially built solution, the algorithm either relaxes v’s bound by a controlled additive term or removes a carefully chosen incident edge, thereby preserving feasibility for the remaining LP.
Re‑optimization – the LP is reconstructed on the reduced graph with the updated constraints, and the process repeats.

Because each iteration fixes at least one variable, the algorithm terminates after at most |E| rounds. The authors prove two crucial properties for the two graph families they consider.

Directed graphs with k‑out‑connectivity: For a root r that must have k internally vertex‑disjoint paths to every other vertex, the iterative rounding yields a subgraph whose total cost is at most 2 · OPT, where OPT is the optimal LP value (hence a 2‑approximation). Moreover, for every vertex v the final degree satisfies
deg_H(v) ≤ 2 b(v) + O(k).
The additive O(k) term stems from the fact that at most O(k) “critical” vertices can become degree‑tight during the process, each contributing at most a constant number of extra incident edges.

Undirected graphs with element‑connectivity: Here the requirement is that for each unordered pair {u, v} there exist k paths that are internally vertex‑disjoint and also edge‑disjoint (element‑connectivity). The same iterative rounding framework gives a 4‑approximation on cost and a degree guarantee
deg_H(v) ≤ 4 b(v) + O(k).
The factor 4 arises because each undirected edge appears twice in the directed biset formulation, effectively doubling the cost bound relative to the directed case.

These results dramatically improve upon the previous O(log k) cost factor and the exponential degree blow‑up. The analysis hinges on structural lemmas about biset families: any minimal fractional solution must contain a biset with a “tight” degree constraint, and the number of such bisets is bounded by a function of k. This combinatorial insight ensures that the additive degree violation remains linear in k, not exponential.

Extensions: The authors demonstrate that the same technique can be adapted to several related problems:

Undirected k‑out‑connectivity (rooted version) – same 4‑approximation and degree bound.
Directed/Undirected k‑connectivity (pairwise requirements) – the algorithm directly yields analogous guarantees.
Rooted k‑connectivity and Subset k‑connectivity – by restricting the biset family to those involving only the designated root or terminal set, the analysis carries over unchanged.

Thus the paper provides a unified framework for a broad class of node‑connectivity design problems with degree restrictions.

Practical considerations: The iterative rounding algorithm requires solving a sequence of LPs, each of which can be handled by modern solvers (e.g., CPLEX, Gurobi, or open‑source alternatives). The rounding step is trivial (checking a threshold), and the degree‑tight handling involves only local modifications. Consequently, the method scales to large networks, and the O(k) additive degree slack is often acceptable in real‑world settings where k is modest (e.g., k = 2–5 for fault tolerance).

Conclusion: By marrying a biset‑based LP relaxation with an elegant iterative rounding process, the authors achieve constant‑factor approximations on cost while keeping degree violations linear in the connectivity requirement. This resolves a long‑standing gap in the literature on degree‑bounded node‑connectivity design, and opens the door to further refinements—such as incorporating additional constraints (capacity, latency) or developing faster combinatorial variants—while preserving the strong theoretical guarantees established in this work.