27-04-2011, 02:17 PM
SUBMITTED BY: -
Manoj Gurjar
[attachment=12935]
ABSTRACT
Wireless multimedia delivery is becoming increasingly more important in today's networks. Unlike wired packet switched networks that suffer from congestion related loss and delay, the wireless networks have to deal with a time varying, error prone, physical channel that in many instances is also severely bandwidth constrained. As such, the solutions needed for wireless video streaming applications are fundamentally different from wired streaming. In this paper, we propose a set of end to end application layer techniques for adaptive video streaming over wireless networks. The adaptation is done both with respect to channel and data. Our approach combines the flexibility and programmability of application layer adaptations, with low delay and bandwidth efficiency of page link layer techniques. Socket level simulations are presented to verify the effectiveness of our approach. Network performance can be increased if the traditionally separated network layers are jointly optimized. Recently, network utility maximization has emerged as a powerful framework for studying such cross-layer issues. In this paper, we review and explain three distinct techniques that can be used to engineer utility-maximizing protocols: primal, dual, and cross decomposition. The techniques suggest layered, but loosely coupled, network architectures and protocols where different resource allocation updates should be run at different time-scales. The decomposition methods are applied to the design of fully distributed protocols for two wireless network technologies: networks with orthogonal channels and network-wide resource constraints, as well as wireless networks where the physical layer uses spatial-reuse time-division multiple access. Numerical examples are included to demonstrate the power of the approach
In this paper, we discuss a cross layer congestion control technique of TCP Reno-2 in wireless networks. In this both TCP layer and PHY layer jointly control congestion. The PHY layer changes transmission power as per the channel condition, interference received and congestion in the network, whereas the TCP layer controls congestion using Reno-2 window based flow control. Our simulations show that the cross layer congestion control technique provides performance improvement in terms of throughput and window size variations.
Keywords: TCP, Reno, Congestion, Optimal, KKT, Lagrangian , Price
1. Introduction
Wireless research has experienced an avalanche of cross layer design techniques. Due to the special characteristics of wireless communication, the traditional layered architecture is coming under closes scrutiny. Cross layer is something that researcher are relying on to solve these problems. There is several different interpretation of cross layer design. It is because researchers have developed cross layer techniques independently. In a broad sense cross layer design refers to protocols that actively use the dependence and communication between different layers to obtain performance gain. But it is also observed that synergy between the performance and implementation complexity is weak. Most of the proposals emphasis on the performance gain keeping implementation concerns unattended. In this report first of all cross layer techniques will be classified into certain categories few protocols of each category will be studied and compared. In the later part of the report current problems and future scope of cross layer technology will be discussed.
1. Congestion in Wireless Network
Wireless networks [1] are inherently limited by battery power and bandwidth constraints. They are characterized by mobility, random changes in connectivity, fluctuations in channel and interference due to neighboring nodes etc. Due to these factors, packet loss of a wireless network is much more than that of a wired network, in which packet loss occurs mainly due to congestion in the network.
Congestion in a network is characterized by delay and packet loss in the network Transport Control protocol (TCP) is used as a reliable transport layer protocol in the traditional best effort (wired) network and deals with congestion effectively The congestion control mechanism of various versions of TCP provides better throughput in an wired network, where the packet loss is mainly due to congestion at various nodes and routers. However, this mechanism may not be suitable in a wireless network, where packet loss due to time-varying nature of channel and interference of other nodes are considerably high Hence, instead of usual congestion control technique; we propose a cross layer technique involving TCP and MAC (Medium Access Control) layer. TCP layer performs the windowing flow control and MAC layer varies transmission power of wireless nodes depending on the channel condition and interference. Our approach consists of (1) Formulation of TCP congestion control mechanism in terms of control system equations (2) Use of transmission power of wireless nodes as a function of cost in an optimization equation (3) Use of optimization techniques to determine the maximum aggregate utility of all the sources, subject to capacity constraints and maximum transmission power [2] of wireless nodes.
2. Congestion Control in Wired Network
Most of the traffic (around 80 %) in the Internet is TCP traffic [4]. TCP’s congestion control in wired network is based on Adaptive Window Management technique. In this technique, congestion window (cwnd) increases or decreases based on packet drops and dupacks. Different versions of TCP use different method to increase/decrease cwnd during
Congestion and are discussed below.
(i) TCP Tahoe: Slow Start and Congestion Avoidance,
(ii) TCP Reno and Reno-2: Fast Retransmit and Fast Recovery, and
(iii) TCP Vegas.
1) TCP Tahoe:
TCP Tahoe [5] assumes losses due to packet corruption are much less probable than losses due to buffer overflows in the network resulting congestion. It uses triple dupacks or timeouts to detect congestion or packet loss in the network. It decreases cwnd size to one (from the current cwnd (size) after detecting a congestion and then increases cwnd size from one to stress (which acts as limit point for exponential increase and is set to half of the cwnd size before experiencing a congestion) an exponential manner in each Round Trip Time (RTT). After reaching stress it increases cwnd linearly in each RTT till next congestion occurs.
2) TCP Reno and Reno-2
Unlike TCP Tahoe, TCP Reno [6] distinguishes between triple dupacks and packet loss (timeout). On packet loss it works similar to TCP Tahoe. But, for a triple dupacks, instead of declaring it as a packet loss and entering the slow start process, it follows fast recovery technique and decreases the cwnd value by half of the current cwnd and then increases linearly till experiencing congestion. Though it is better than TCP Tahoe for dealing with single packet loss and dupacks, it is not good when multiple packets are lost within one RTT. This problem is solved in its newer version called TCP Reno- 2. In TCP Reno-2 [3], the cwnd value is not decremented for every packet loss, rather is decremented in an intelligent manner. Cwnd is decremented by half only when one or more than one packet loss occurs in an RTT. In this technical report, we study congestion control in TCP Reno-2 in wireless network.
3) TCP Vegas:
In TCP Vegas [7], cwnd size is increased or decreased depending on the difference of ratio of current window size, propagation delay and queuing delay. This congestion control mechanism is similar to a sliding windows protocol.
3 Cross layer Design Framework
Fig.1 System diagram for the cross-layer design framework
As illustrated in Fig. adaptive page link layer techniques will be used to adjust the capacity of individual wireless links to support delay-constrained traffic, possibly in multiple service classes; dynamic capacity assignment in the media access (MAC) layer will optimally allocate resources among various traffic flows; a congestion-optimized routing algorithm will provide multiple paths to real-time media streams; finally at the transport and application level, intelligent packet scheduling and error-resilient audio/video coding will be optimized for low-latency delivery over ad-hoc wireless networks.
The proposed framework will integrate the above components in a dynamic and iterative fashion. It allows the exchange of relevant information such as page link capacities, traffic flows, packet deadlines and rate-distortion preamble of the source data across the entire protocol stack
2. Classification
Cross layer techniques can be divided into following broad categories:
1. Rate Adaptation :
Rate adaptation is an inevitable feature of wireless links. In traditional layer based archi-
tecture , physical layer details are abstracted out of mac layer. So mac layer can decide its
transmission rate only on basis of successful packet reception rate. This prevents mac layer from taking quick rate change decisions. Cross layer methods enables mac layer to collect physical layer information. On basis of this information dynamic rate adaptation is achieved. Following techniques of rate adaptation are studied and compared in details.
CHARM
FARA
SoftRate
2. Network Coding :
Network coding involves mixing up of data in the network. It may be of two types. First is about encoding data at sender. This involves packing up data for multiple receivers in the same packet. This increases information flow through the network. Other type is collecting data from multiple receivers and combining them to get a meaningful data packet. We will study three protocols. COPE is an example of first kind. PPR involves second type of procedure. Whereas MIXIT is a combination of both.
3. Multiradio Enhancements :
Presence of multiple physical layer (radio) is one of most common method for cross layer. Upper layer controls different physical layers and uses the data received by these layer. Two protocols can be defined under this category.
MRD
2P Ma
4. Directional Antenna :
In this type of cross layer , the physical layer is characterized by the presence of directional antennas. These antennas can be directed or beamed towards specific target to create P2P abstraction. This prevents the classical hidden node problem in wireless network. The orientation of these antennas are controlled by software or MAC layer. We have the following protocols to study
Mobisteer
DIRC
5. Sniff and Retransmit :
In these category of protocol , lower layer sniffs data from upper layer and caches it to overcome the shortcomings of the upper layer. Few examples of this kind of protocol are
Snoop
PRO
