Performance of distributed mechanisms for flow admission in wireless adhoc networks

Given a wireless network where some pairs of communication links interfere with each other, we study sufficient conditions for determining whether a given set of minimum bandwidth quality-of-service (QoS) requirements can be satisfied. We are especially interested in algorithms which have low communication overhead and low processing complexity. The interference in the network is modeled using a conflict graph whose vertices correspond to the communication links in the network. Two links are adjacent in this graph if and only if they interfere with each other due to being in the same vicinity and hence cannot be simultaneously active. The problem of scheduling the transmission of the various links is then essentially a fractional, weighted vertex coloring problem, for which upper bounds on the fractional chromatic number are sought using only localized information. We recall some distributed algorithms for this problem, and then assess their worst-case performance. Our results on this fundamental problem imply that for some well known classes of networks and interference models, the performance of these distributed algorithms is within a bounded factor away from that of an optimal, centralized algorithm. The performance bounds are simple expressions in terms of graph invariants. It is seen that the induced star number of a network plays an important role in the design and performance of such networks.

💡 Research Summary

**
The paper addresses the flow‑admission problem in wireless ad‑hoc networks where links interfere with each other. Interference is modeled by a conflict graph G₍C₎ whose vertices represent communication links and edges indicate pairs that cannot be active simultaneously. Each link ℓ has a demand f(ℓ) (bits per second) and a maximum transmission capacity Cℓ; the total shared bandwidth of the medium is C. Satisfying all demands amounts to finding a schedule that activates independent sets of G₍C₎ for appropriate fractions of time. This scheduling problem is equivalent to computing the fractional chromatic number of a weighted graph, which is NP‑hard in general.

Because a centralized optimal solution is impractical for large, dynamic networks, the authors study distributed admission‑control mechanisms that rely only on local information. Four families of sufficient conditions are examined:

1. Row constraints – for each link ℓ, the sum of demands of ℓ and all its neighbors must not exceed C.
2. Degree constraints – each link’s demand multiplied by (its degree + 1) must be ≤ C.
3. Mixed constraints – the tighter of the row and degree constraints is used.
4. Scaled‑clique constraints – for every clique K in G₍C₎, the total demand of K divided by |K| must be ≤ C.

Each condition yields an upper bound on the required fractional chromatic number, and therefore a guarantee that if the bound holds the demand vector is feasible. The paper’s main contribution is a rigorous worst‑case performance analysis of these distributed algorithms. The authors introduce the induced star number σ(G₍C₎) – the maximum number of leaves in an induced star subgraph – and prove that the worst‑case approximation ratio of the row constraints is exactly σ(G₍C₎). Consequently, for any network where σ(G₍C₎) is bounded, the row‑based distributed algorithm is guaranteed to be within that factor of the optimal centralized solution.

The authors apply this framework to several well‑studied network classes:

• Unit‑disk graphs (nodes with equal transmission range). It is known that σ ≤ 5 for these graphs; thus the row‑constraint algorithm is at most a factor‑5 away from optimal. Earlier work showed that scaled‑clique constraints achieve a 2.1‑approximation, but require more global information.
• Primary‑interference model (a node cannot participate in more than one transmission at a time). Here the conflict graph is essentially a line graph, giving σ ≤ 3, so the row‑based method is within a factor‑3 of optimal.
• Additional examples (grid topologies, random geometric graphs) are examined through simulation, confirming that average performance is often much better than the worst‑case bound (typically 1.5–2×).

From an implementation perspective, all four distributed conditions can be evaluated with only O(Δ) computation and communication per node, where Δ is the maximum degree of the conflict graph. Nodes exchange their own demand and possibly degree or clique information with immediate neighbors, avoiding any need for a central controller or network‑wide state. The paper also discusses how to translate a fractional schedule into a concrete time‑slot schedule using rational approximations, ensuring that the theoretical guarantees can be realized in practice.

In summary, the work provides a clear theoretical link between a simple graph invariant (the induced star number) and the performance of low‑overhead distributed admission‑control algorithms. It shows that for many realistic wireless topologies the loss relative to an optimal centralized scheduler is bounded by a small constant, making these distributed mechanisms attractive for real‑world ad‑hoc and sensor networks where scalability and low latency are essential.