Consider a wireless network where each communication link has a minimum bandwidth quality-of-service requirement. Certain pairs of wireless links interfere with each other due to being in the same vicinity, and this interference is modeled by a conflict graph. Given the conflict graph and link bandwidth requirements, the objective is to determine, using only localized information, whether the demands of all the links can be satisfied. At one extreme, each node knows the demands of only its neighbors; at the other extreme, there exists an optimal, centralized scheduler that has global information. The present work interpolates between these two extremes by quantifying the tradeoff between the degree of decentralization and the performance of the distributed algorithm. This open problem is resolved for the primary interference model, and the following general result is obtained: if each node knows the demands of all links in a ball of radius $d$ centered at the node, then there is a distributed algorithm whose performance is away from that of an optimal, centralized algorithm by a factor of at most $(2d+3)/(2d+2)$. The tradeoff between performance and complexity of the distributed algorithm is also analyzed. It is shown that for line networks under the protocol interference model, the row constraints are a factor of at most $3$ away from optimal. Both bounds are best possible.

中文翻译：

---

无线自组织网络中QoS分布式算法的性能保证

考虑一个无线网络，其中每个通信链路都有最低带宽服务质量要求。由于位于同一附近，某些无线链路对会相互干扰，这种干扰由冲突图建模。给定冲突图和链路带宽要求，目标是仅使用本地化信息确定是否可以满足所有链路的需求。在一种极端情况下，每个节点只知道其邻居的需求；在另一个极端，存在具有全局信息的最优、集中式调度器。目前的工作通过量化去中心化程度和分布式算法性能之间的权衡，在这两个极端之间进行插值。主要干扰模型解决了这个未解决的问题，得到以下一般结果：如果每个节点都知道以该节点为中心的半径为 $d$ 的球内所有链路的需求，则存在一个分布式算法，其性能与最优的集中式算法的性能相差最多为 $(2d+3)/(2d+2)$ 的因子。还分析了分布式算法的性能和复杂度之间的权衡。结果表明，对于协议干扰模型下的线路网络，行约束距离最优值最多为$3$。两个边界都是最好的。还分析了分布式算法的性能和复杂度之间的权衡。结果表明，对于协议干扰模型下的线路网络，行约束距离最优值最多为$3$。两个边界都是最好的。还分析了分布式算法的性能和复杂度之间的权衡。结果表明，对于协议干扰模型下的线路网络，行约束距离最优值最多为$3$。两个边界都是最好的。