SEÑAL DE PUBLICIDAD COMERCIAL

• B. R´es, L. Barreto, and S. Sargento, ”rt-winf : Real time wireless inference mecha- nism.” in IEEE Globecom 2010 Workshop on Mobile Computing and Emerging Com- munication Networks (MCECN 2010), Miami, Florida, USA, 2010.

• L. Barreto and S. Sargento, ”XCP-winf and RCP-winf : Congestion control tech- niques for wireless mesh networks.” in IEEE International Conference on Commu- nications (ICC 2011), Kyoto, Japan, 2011.

1.4.2

National Proceedings with Independent Review

• L. Barreto and S. Sargento,”How Real-time Bandwidth Inference Improves the Con- gestion Control in Wireless Mesh Networks.” In Actas da 10 Conferˆencia sobre Redes de Computadores (CRC’2004), Braga, Portugal, November 2010.

1.4.3

Pending

• L. Barreto and S. Sargento, ”XCP-winf and RCP-winf : A new approach for wireless congestion control.” submitted to Springer AD HOC NETWORKS journal, Novem- ber 2011.

• L. Barreto and S. Sargento, ”WE TCP-AP : Wireless Enhanced TCP-AP.” submitted to Springer COMNET - COMPUTER NETWORKS journal, January 2012.

1.5

Thesis Organization

The remainder of this Thesis is organized as follows. Chapter 2 presents the state of the art regarding the main and major available bandwidth and link capacity estimation techniques used in wired and wireless networks. This chapter also discusses and presents the related work on congestion control mechanisms, including both rate based congestion control mechanisms and Additive Increase Multiplicative Decrease (AIMD) congestion con- trol mechanisms. Chapter 2 ends with the discussion of the principle models and research carried out in collision probability and cross layer design.

In Chapter 3 we present a performance study and analysis of congestion control proto- cols. We establish a set of strong evaluation parameters that will be used in all congestion control performance evaluation along this Thesis. We analyze the performance of some experimental AIMD based congestion control protocols against TCP, WCP and TCP-AP. Rate based congestion control protocols are also analyzed against TCP. The proposals cho- sen are XCP, RCP and XCP-b. This Chapter allows to understand that much work is still needed in the definition of efficient congestion control mechanisms for wireless multi-hop networks.

Chapter 4 introduces a new available bandwidth and link capacity estimation technique, called rt-Winf. We provide a detailed description, implementation, and a performance

study of rt-Winf against other popular estimation mechanisms. The performance study is done using the ns-2 simulator and the CMU wireless emulator [24].

In Chapter 5, we discuss the design, implementation, and evaluation of two new exper- imental congestion control mechanisms for wireless environments, XCP-Winf and RCP- Winf. We demonstrate that the use of MAC layer information improves congestion control performance that results in overall improved network performance. The results and con- clusions obtained also motivate the need for cross layer approaches to wireless congestion control. In the comparison of XCP-Winf and RCP-Winf, we evaluate the metrics used in Chapter 3, including a set of measurements that allowed us to verify TCP friendliness.

In Chapter 6 we present a new congestion control mechanism based in TCP-AP, WE TCP-AP. TCP-AP is a TCP based protocol, which was developed to improve TCP per- formance in wireless ad-hoc networks. We provide a performance evaluation of TCP-AP against other wireless congestion control techniques. With the performance results and the identification of the factors that limit TCP-AP performance, we then, focus on proposing enhancements to be applied into the original TCP-AP design. The Chapter ends with a performance evaluation of the improved TCP-AP protocol, the WE TCP-AP.

In Chapter 7 we summarize the contributions of this Thesis, discuss the limitations of this work, present possible future directions of this work, and outline the concluding remarks.

Chapter 2

Related Work

2.1

Introduction

Congestion control has been a high debated and researched issue in wired and wireless networks. A large number of protocols and mechanisms have been proposed with the main objective to improve congestion control, specially in wireless environments. Some of the proposals try to enhance Transport Control Protocol (TCP) behavior, incorporating new congestion parameters and algorithms for wireless networks. These proposals rely on TCP modifications that address specific inefficiencies of TCP itself, or rely on routing or link layer improvements aimed at improving TCP performance, while keeping unchanged the Additive Increase Multiplicative Increase (AIMD) behavior of TCP. We can classify these proposals as AIMD based congestion control mechanisms. Some of its examples are TCP-Feedback (TCP-F), TCP-BuS or the Wireless Control Protocol (WCP).

Other mechanisms, however, focused in using rate control feedback between receiver and sender, to improve network performance. We can classify these proposals as rate based congestion control mechanisms. The eXplicit Control Protocol (XCP) and the Rate Control Protocol (RCP) are two examples of rate based congestion control mechanisms developed for wired networks. Some new proposals like the XCP-blind (XCP-b) and the Wireless RCP (WRCP) were defined for wireless congestion control.

Recent research on congestion control tries to use the AIMD process of TCP together with a rate based congestion control scheme. These proposals attempt to benefit from the best of the two worlds. We can classify these new approaches as hybrid congestion control mechanisms. Some examples of hybrid congestion control schemes, specially developed for wireless environments, are TCP with Adaptive Pacing (TCP-AP) and WCP Capacity based (WCPCap).

In this chapter, we first discuss the main operating principles of TCP and its problems to deal with wireless environments. Then, we describe some of the congestion control protocols specially developed for wireless networks.

When discussing wireless networks congestion control, link capacity and available band- width are two important characteristics that can be used for congestion control improve-

ment, allowing a better medium usage and a better congestion control. Link capacity and available bandwidth estimation have been widely studied, specially in wired networks. However, new efforts have been made in developing estimation mechanisms that can be applied in the context of wireless networks. We then discuss in this chapter some of the most used mechanisms for capacity and available bandwidth estimation.

For higher accuracy, both link capacity and available bandwidth should be inferred at the Medium Access Control (MAC) layer. Then, to optimize transport layer behavior and improve the overall network performance cross layer information, enabling layer interaction is fundamental. Other aspect that must be taken into consideration when the communi- cating nodes are mobile is that, due to the nature of mobile communications, collision probability is increased. This chapter then introduces some of the more relevant work that has been addressed in those fields.

As a resume, this chapter main aim is to provide the fundamental knowledge of mecha- nisms and protocols that will allow the definition of our new proposed estimation techniques and congestion control mechanisms.

This chapter is organized as follows. Section 2.2 describes the major available band- width and link capacity estimation techniques, categorized into active measurement tools (section 2.2.1) and passive measurement tools (section 2.2.2). Then, section 2.3 discusses the main congestion control schemes. Section 2.3.1 introduces the TCP, the most used congestion control mechanism. Next, section 2.3.1.1 presents a set of TCP’s limitations and issues in wireless networks. Then, section 2.3.2 presents congestion control approaches that use TCP’s AIMD scheme, and section 2.3.3 introduces the mechanisms that use a rate control scheme for congestion control. Section 2.3.3.1 presents the main operating principles of XCP and some XCP-based congestion control mechanisms for wireless net- works. Section 2.3.3.2 introduces the main operating principles of RCP and a RCP-based congestion control mechanisms for wireless environments. Then, section 2.3.4 discusses the congestion control mechanisms that are both AIMD and rate based. Section 2.4 presents the most important research on collision probability in wireless networks dynamics, and finally, section 2.5 describes the most used models and some of the research carried out in cross layer design.