FREE ELECTRONIC LIBRARY - Theses, dissertations, documentation

Pages:     | 1 |   ...   | 20 | 21 || 23 | 24 |   ...   | 39 |

«Carles Gómez Josep Paradells José E. Caballero Edita: Fundación Vodafone España Autores: Carles Gómez Montenegro* Universitat Politècnica de ...»

-- [ Page 22 ] --

In effect, DTC adapts the main idea behind Snoop [25] for WSNs. Snoop was designed as a TCP proxy placed at the base station for wireless cellular networks. Snoop caches TCP segments transmitted in the downlink (i.e. to the mobile client) and maintains local timers so as to locally retransmit the lost data segments. In this way, losses in the wireless link can be hidden from the sender and its retransmission and congestion control mechanisms can be avoided. While in DTC the described mechanisms may be present in various intermediate nodes between sender and receiver, Snoop resides only in the base station, that is, a single element between sender and receiver.

–  –  –

UDP is a lightweight protocol that mainly adds source and destination port information to IP datagrams. It offers neither reliability nor congestion control. For this reason, UDP can be used in WSNs for applications that do not require reliability in contexts of low congestion, while offering interoperability with IP networks (e.g. the Internet). Furthermore, as an alternative to the use of TCP, UDP has been used for WSN devices running IP, augmented with acknowledged and numbered packet delivery. This approach provides reliability and may be simpler than adapting TCP21. UDP can also be suitable when application layer protocols offer reliability.

8.1.5. Use of transport layer proxies

A proxy is an entity placed between the endpoints of a communication and is designed for improving degraded performance caused by the nature of specific environments [24].

When the nodes of a WSN run a special transport layer protocol (e.g. a simplified version of a traditional transport protocol or a new protocol specifically designed for WSNs), the communication with devices in another network (e.g. hosts in the Internet) may benefit from the use of a proxy operating at the transport layer. Such proxies are referred to as transport layer proxies.

A proxy may either be centralized (e.g. in a gateway that connects the WSN with the Internet) or distributed between two or more devices (e.g. a gateway and the WSN nodes). Some types of proxies break the end-to-end principle, which is one of the fundamentals of the Internet architecture, but it has been shown that they can dramatically enhance the performance of Internet communications over limited environments [24]. On the other hand, a transport layer proxy may enable the sensors within a WSN to be seen as devices fully compliant with protocols such as TCP, while the sensors may run a simplified version of those protocols or completely different ones.

As an example, the Compact Application Protocol (CAP) proposes running the ZigBee Application Protocol on top of UDP/IP networks (see Section 13.3.3, Figure 13).

–  –  –

8.2. Protocols for congestion control in WSNs Congestion may occur in WSNs due to two main causes: i) the packet arrival rate at a node exceeding the node’s service rate, and ii) PHY/MAC layer phenomena such as interference, errors and contention. This section presents various protocols designed for congestion control in WSNs. They basically differ in congestion detection, congestion notification and congestion mitigation mechanisms.

In the following, it is assumed that in a tree configuration of a sink-based WSN (as in Fig. 8.1) the downstream direction means that source node is the sink node and destination node is the sensor node, while the upstream direction refers to the reverse direction. Thus, downstream traffic may be one-to-many (e.g. the sink node disseminates a request or an update to more than one sensor node) or one-to-one (if a message is transmitted from the sink node to a specific sensor node), while in the upstream direction traffic is generally many-to-one (e.g. various sensors of the same type transmit the data they collect to the sink).

Typically, the protocols presented in this section are defined to overcome congestion issues in the upstream direction of a tree WSN topology.

In some WSN scenarios (e.g. control networks for building automation), one-to-one communication between arbitrary nodes may be a significant fraction of the traffic. The protocols presented in this section do not assume such scenarios.

Fig. 8.1. Downstream vs. upstream traffic directions in a typical WSN

–  –  –

8.2.1. CODA Congestion Detection and Avoidance (CODA) [10] is a congestion control scheme for WSNs. In particular, CODA was designed for event-driven WSNs, where congestion may take place once a certain event has been detected.

CODA is composed of three mechanisms (coping with congestion detection, notification and mitigation features respectively): i) receiver-based congestion detection, ii) open-loop hop-by-hop backpressure messages, and iii) closed-loop multi-source regulation.

i) Queue occupancy may be used as a measurement to detect congestion. However, it may only be accurate when a queue is completely empty or nearly overflowing [10]. For this reason, CODA uses a channel load measurement and compares it with a certain level.

In order to minimize the energy cost of such approach, this task is carried out only at certain moments. More specifically, CODA assumes the use of a CSMA MAC and takes advantage of the Clear Channel Assessment procedure of this MAC, which requires a sensor node to put the radio in receiver mode and listen to the medium before transmitting, and hence incurs no extra cost.

ii) Once a sensor node detects congestion, it broadcasts backpressure messages to its neighbours. The sensor node may continue transmitting these messages every certain period if congestion persists. Backpressure messages are transmitted toward the source. When a node receives a backpressure message, it decides whether or not to forward the message towards the source, according to its local conditions. On the other hand, this sensor node may perform actions according to the local congestion policy, such as dropping incoming data packets or Adaptive Increase Multiplicative Decrease (AIMD) rate adjustment.

iii) When the event detection rate of a sensor node is greater than a certain threshold, the sensor node activates the so-called ‘regulate bit’ in the event packets that it transmits to the sink node. When the sink node receives a packet with the regulate bit set, it sends ACKs to all the sources related with that event. Reception of such ACKs by the sources allows them to maintain the same transmission rate. If the sources do not receive the ACKs,

–  –  –

they reduce their transmission rate. ACKs may be lost in the network due to congestion. On the other hand, the sink node may stop transmitting ACKs according to its own local congestion measurements.

8.2.2. Fusion Fusion [9] is a congestion control scheme for WSNs that combines three techniques for detection, notification and mitigation of congestion: i) hopby-hop flow control, ii) rate limiting, and iii) a prioritized MAC.

i) The goal of hop-by-hop flow control in Fusion is to avoid packet transmission if packets are to be discarded due to congestion in the paths towards the destination. This mechanism comprises two significant components: congestion detection and congestion mitigation. Congestion can be detected by monitoring the occupancy of the sensor node queue or by measuring the utilization of the channel. In Fusion, once congestion is detected, a congestion bit is set in the header of data packets (which are transmitted to the sink). Congestion mitigation is performed as follows: Let us assume that a parent is the next hop of a given sensor node in the path towards the sink; this sensor node is a child of the parent node. If a sensor node overhears a packet from its parent with the congestion bit set, then it stops forwarding data. If this sensor node and other children of that parent become congested in turn, they are allowed to transmit once (with the congestion bit set) so that their own children can learn that they are congested.

ii) The rate limiting mechanism in Fusion is based on the use of a token bucket to regulate the rate of each sensor node. This mechanism is described next. Every time a sensor node hears that its parent forwards N packets, its token count is incremented by one. The sensor node can only transmit when there are tokens available, and each transmission decreases the token count by one. In this way, each sensor node sends at the same rate as each of its descendants. This mechanism reduces the probability of congestion points occurring as well as increasing fairness.

iii) Finally, Fusion uses a MAC layer technique that prioritizes the transmission of congested nodes. In particular, a CSMA MAC is assumed, and the

–  –  –

back-off window of congested nodes is one fourth the size of a non-congested sensor’s back-off window. With this technique, the probability of winning the contention period of a congested sensor node increases.

–  –  –

The algorithm for Congestion Control and Fairness (CCF) [16] also aims at eliminating congestion in a WSN while assuring fair packet delivery to a sink node. It operates at the transport layer on top of any MAC layer. Note that this is not always so, as is the case in CODA or Fusion (which run on top of a CSMA MAC layer).

The algorithm deals with two types of congestion. The first occurs when several sensor nodes within the same range wish to transmit at the same time. To reduce this kind of congestion, jitter (i.e. a small random delay variation before transmission) is introduced. Secondly, queue occupancy is monitored in order to detect congestion within a specific sensor node. If the occupancy of the queue that holds packets before they are transmitted exceeds a certain threshold (before the queue becomes full), the downstream sensor nodes are informed in order to reduce their transmission rates.

To implement the aforementioned ideas, the algorithm measures the actual rate r at which packets can be transmitted from a given sensor node, which is the inverse of the total service time required by the transport layer. This service time includes all the congestion effects that a node may suffer. It then calculates the maximum rate at which the nodes of its complete sub-tree can transmit, rdata, as the rate r divided by the number of devices in the sub-tree. Every sensor node also calculates the rate rdata,qfull, at which its queues are about to overflow, as well as the sensor node’s data generation rate, rdata,parent. parent. The children nodes are then periodically informed of the maximum transmission rate they are allowed to use, which is set to min{rdata, rdata,qfull, rdata,parent.}. Note that with this algorithm, the rate assignment for each node is fair.

Sensors Everywhere

The mechanism described constitutes a closed loop that is in constant oscillation. Fig. 8.2 illustrates an example. If the shaded sensor node (Fig. 8.2.a)) experiences congestion, it informs its children nodes in order to reduce their transmission rate (Fig. 8.2.b)). This information is propagated throughout the network, and because smaller rates are used, node queues become empty and the time to transmit a packet successfully decreases (Fig. 8.2.c)). The shaded node disseminates the new (increased) rate (Fig. 8.2.d)) again, which in turn may cause congestion and thus the cycle repeats.

Fig. 8.2. CCF operation (adapted from [16])

Because rate updates reach the nodes with a time proportional to the hop count from the parent node (e.g. the shaded node in the example), sensor nodes of different depths experience phase shifts in their transmissions, which are equivalent to the jitter that it was desirable to introduce, as explained above.

–  –  –

8.2.4. PCCP Priority-based Congestion Control Protocol (PCCP) [12] aims at controlling congestion in a WSN, while addressing the fact that nodes may have different priorities depending on their location and/or function. PCCP is composed of three main components: i) Intelligent Congestion Detection (ICD), ii) Implicit Congestion Notification (ICN), and iii) Priority-based Rate Adjustment (PRA).

ICD detects congestion according to packet inter-arrival and packet service times at the MAC layer of a node. Packet inter-arrival is defined as the time between two consecutive arriving packets, which can either be generated by a different node, required to be forwarded by this node, or alternatively can be packets created by this node. The packet service time is defined as the time difference between the instant at which the packet arrives at the MAC layer and the instant at which the last bit of the packet is transmitted. PCCP uses a parameter called congestion degree, which is the ratio of average packet service time over average packet inter-arrival time. If the congestion degree is greater than one, the node experiences congestion.

In order to propagate congestion information, PCCP uses ICN, which is based on piggybacking congestion information in the header of data packets. The transmission of these packets can be overheard by the nodes within the range of the corresponding sender. Hence, the children of a node can know when their parent is congested.

Finally, PRA requires the introduction of a scheduler with two subqueues between the network layer and MAC layer (see Fig. 8.3). One sub-queue is for traffic generated in the node and the other is for transit traffic. On the basis of three priority index values, the scheduler gives the appropriate weight to each sub-queue and the scheduling rate is calculated in order to avoid or mitigate congestion. This calculation uses the congestion degree and priority index values for accurate rate adjustment.

–  –  –

Simulation results show that PCCP outperforms CCF in terms of throughput. This happens because CCF does not take into account the packet interarrival time (recall that CCF is based on measuring the packet service time).

When errors take place in the wireless links, inter-arrival time increases, leading to channel under-utilization. CCF erroneously treats such under-utilized links as congested links.

8.2.5. Siphon

Pages:     | 1 |   ...   | 20 | 21 || 23 | 24 |   ...   | 39 |

Similar works:

«Supporting Aboriginal and Torres Strait Islander children in kindergarten July 2005 This document is managed by the Department of Education and Early Childhood Development, Victoria (as of 27 August 2007) Supporting Aboriginal and Torres Strait Islander children in kindergarten July 2005 This document is managed by the Department of Education and Early Childhood Development, Victoria (as of 27 August 2007) Copyright State of Victoria, Department of Human Services, 2005 This publication is...»


«France Case Study Brussels, November 2013 France Case Study Brussels, November 2013 Table of contents Executive Summary 7 1 Description of COSMED 9 1.1 Background 9 1.2 Who is COSMED 9 1.3 Committees and Working Groups 10 1.4 Regulatory Monitoring 11 1.5 Buying Group 11 1.6 Training 12 2 Internationalisation dynamics 14 2.1 COSMED 14 2.2 COSMED business members 14 2.3 How could the EC help to face the internationalisation challenges 15 3 Impact of selected business network 17 3.1 Value added of...»

«tagliatella castellana tagliatella castellana La Tagliatella Castellana in Madrid Book a table at La Tagliatella Castellana in Madrid. Find restaurant reviews, menu, prices, and hours of operation for La Tagliatella Castellana on TheFork. Tagliatelle Castellana Le ricette di Tagliatelle Castellana: Fate dorare la cipolla e l'aglio tritati finemente in una noce di burro facendo attenzione che non s La Tagliatella, Madrid Paseo de la Reserva La Tagliatella, Madrid en TripAdvisor: Consulta 59...»

«An infra-red finger tracking system used in the assessment and remediation of “graph-as-picture” misconceptions Grecia Garcia Garcia1 and Richard Cox2 University of Sussex, Brighton, BN19RH, UK gg44@sussex.ac.uk The University of Edinburgh, Edinburgh, EH8 9AB, UK rcox@inf.ed.ac.uk Abstract. The workshop presentation will describe a specialized application of Lee’s “Wiimote Whiteboard” [7] an infra-red camera based tracking system which uses the Nintendo Wii wireless remote control...»


«Congratulations on being selected for the Baseball Factory Spring Training in Arizona! Baseball Factory scouts select players from across the country for this event based on their performance at an Under Armour National Tryout, Premium Video Program or other Baseball Factory events. We are very excited to work with you and help take your game to the next level. As you are aware, this event will take place from Friday, January 1 5th Monday, January 1 8th. This packet has been created as a...»

«Preliminary Please do not cite. Comments welcome.Cross-regional variations in offshore outsourcing choices: evidence from firm-level data Eiichi Tomiura*, Banri Ito§and Ryuhei Wakasugi# June 11, 2008 To be presented at the University of Nottingham for the GEP Conference on Offshoring Abstract In offshoring, a firm chooses outsourcing to independent suppliers or in-sourcing from their FDI subsidiaries. This paper empirically examines how the factor intensity is related with the firm’s...»

«1 ECU Family Therapy Clinic Policies and Procedures Manual East Carolina University Latest Edition 2011-2012 POLICIES AND PROCEDURES REGARDING THE FAMILY THERAPY CLINIC Purpose of Manual This manual contains the policies and procedures of the Family Therapy Clinic at East Carolina University (ECU). The manual was developed for use by students, faculty, and administrators associated with the master‟s and doctoral programs, as well as, therapists, clinical supervisors, and staff working at the...»

«1 The Problem of Possibilia Are there, in addition to the various actual objects that make up the world, various possible objects? Are there merely possible people, for example, or merely possible electrons, or even merely possible kinds? We certainly talk as if there were such things. Given a particular sperm and egg, I may wonder whether that particular child which would result from their union would have blue eyes. But if the sperm and egg are never in fact brought together, then there is no...»

«Mutations and Genetic Variability 1. What is occurring in the diagram below? A. Sister chromatids are separating. B. Alleles are independently assorting. C. Genes are replicating. D. Segments of DNA are crossing over.2. The chart below shows the codons that make up the genetic code and the sequence of nucleotides that corresponds to them. A mistake during DNA replication leads to a mutation in the nucleotide sequence shown below. This mutation results from the insertion of two nucleotides into...»

 PANEM Using the internet, search on Google (or a similar search engine) the meaning of the word “Panem.” Then answer the following questions: 1. What is the full Latin phrase for “Panem”? _2. What does it mean (the literal translation)? _ 3. IN YOUR OWN WORDS, describe how this term came into use: 4. Now that you know its origin and meaning, explain why you think author Suzanne Collins chose the name Panem for the country in which...»

<<  HOME   |    CONTACTS
2016 www.theses.xlibx.info - Theses, dissertations, documentation

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.