Wireless mesh network performance: throughput and fairness

无线网状网络是一种新的、颠覆性的技术，为农村和城市用户提供高速的最后一英里接入。例如，多伦多水电公司宣布，他们将建造加拿大最大的WiFi网络，为多伦多居民提供低成本的互联网接入。虽然这种网络的核心组成部分（无线电发射机和基本网络协议）已经存在，但仍然存在许多问题。在过去的几年里，我的小组一直致力于一系列多跳无线网络主题。我们与渥太华的一家小型初创公司合作，建立了加拿大为数不多的无线网状测试平台之一。在拟议的研究中，我们将在此基础上解决核心挑战之一：为最终用户提供公平和高带宽（即，高吞吐量）接入无线网格并通过它接入因特网。1）吞吐量：无线链路通常是带宽受限的并且易受干扰。大多数MAC和路由协议未能完全利用网络资源，导致端到端吞吐量低于可能的总吞吐量。一些分析模型存在，以确定最大可实现的吞吐率，但不能直接应用，由于其集中的性质和高计算复杂性。我们将开发近似这样的模型，并使用它们来获得分布式协议，实现的关键思想，以最大限度地提高端到端的吞吐量。2）公平性：用户应该实现他们的“公平”共享网络资源时，访问网格。然而，使用当前的协议，带宽共享是非常不均匀的，并且依赖于位置，用户更接近固定基础设施，并且具有较低RTT的连接从网络中获得更好的服务。这两个主题是内在联系的：确保公平性将以总吞吐量为代价。我们将探索这些主题使用分析模型，模拟研究，并通过在我们的测试平台的实施。我们的目标是开发更好的网络协议，从MAC到传输层，并通过模拟和测试平台进行评估。

Wireless mesh networks are a new, disruptive technology to provide high-speed last-mile access to rural and urban users. Toronto Hydro, for example, announced that they will build Canada's largest WiFi mesh to provide low-cost Internet access to Toronto residents. While the core building blocks of such networks (radio transmitters and basic networking protocols) exist, many problems remain. In the past few years, my group has worked on a range of multi-hop wireless network topics. In conjunction with a small Ottawa startup company, we built one of the few wireless mesh testbeds in Canada. In the proposed research, we will build on this basis to address one of the core challenges: providing end-users with fair and high-bandwidth (i.e., high throughput) access to the wireless mesh and through it to the Internet.1) Throughput: wireless links are typically bandwidth-constrained and subject to interference. Most MAC and routing protocols fail to completely utilize the network resources, resulting in lower than possible aggregate end-to-end throughput. A few analytical models exist to determine the maximally achievable throughput rates, but cannot be applied directly due to their centralized nature and high computational complexity. We will develop approximations to such models and use them to derive distributed protocols that implement the key ideas to maximize end-to-end throughput.2) Fairness: users should achieve their "fair" share of network resources when accessing the mesh. Yet using current protocols, bandwidth sharing is very uneven and location-dependent, with users closer to the fixed infrastructure and with connections with lower RTT getting better service from the network.The two topics are intrinsically linked: ensuring fairness will come at an aggregate throughput penalty. We will explore these topics using analytical model, simulation studies, and through implementation in our testbed. The goal is to develop better network protocols from the MAC to the transport layer and to evaluate them through simulation and in our testbed.