Distributed point coordination function for IEEE 802.11 wireless ad hoc networks

Ad Hoc Networks 10 (2012) 536–551

Contents lists available at SciVerse ScienceDirect

Ad Hoc Networks journal homepage: www.elsevier.com/locate/adhoc

Distributed point coordination function for IEEE 802.11 wireless ad hoc networks J. Alonso-Zárate a,⇑, C. Crespo b, Ch. Skianis c, L. Alonso b, Ch. Verikoukis a a

Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Castelldefels, Barcelona, Spain Department of Signal Theory and Communications, Universitat Politècnica de Catalunya (UPC-EETAC), Castelldefels, Barcelona, Spain c University of the Aegean, Greece b

a r t i c l e

i n f o

Article history: Received 16 February 2011 Received in revised form 28 July 2011 Accepted 8 September 2011 Available online 8 October 2011 Keywords: MAC DCF PCF IEEE 802.11 DPCF DQCA DQMAN

a b s t r a c t The Distributed Point Coordination Function (DPCF) is presented in this paper as a novel Medium Access Control Protocol (MAC) for wireless ad hoc networks. DPCF extends the operation of the Point Coordination Function (PCF) defined in the IEEE 802.11 Standard to operate over wireless networks without infrastructure. In PCF, a central point coordinator polls the users to get access to the channel and data collisions are completely avoided, thus yielding high performance. In order to extend its high performance to networks without infrastructure, the DPCF is proposed in this paper as a combination of the Distributed Coordination Function (DCF) and the PCF. The general idea is to combine a dynamic, temporary, and spontaneous clustering mechanism based on DCF with the execution of PCF within each cluster. The backwards compatibility of DPCF with legacy 802.11 networks is also assessed in this paper. Comprehensive computer-based simulations demonstrate the high performance of this new protocol in both single-hop and multi-hop networks. Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction The Medium Access Control (MAC) layer defines the set of rules that the mobile stations of a wireless network should obey to get access to the radio channel in an efficient manner. The IEEE 802.11 Standard, which defines the MAC and physical (PHY) layers for Wireless Local Area Networks (WLANs), was first released in 1999 and reissued later in 2007 grouping some of the subsequent amendments [1]. Since the first release, it has been widely accepted as the most commonly used interface for short-range wireless connectivity. The Distributed Coordination Function (DCF) is defined in this standard as the mandatory access method for any compliant device. This access method is based on Carrier Sensing Multiple Access (CSMA), i.e., listen before transmit, in combination with a Binary Exponential Backoff (BEB) ⇑ Corresponding author. Tel.: +34 936452900; fax: +34 936452901. E-mail address: [email protected] (J. Alonso-Zárate). 1570-8705/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.adhoc.2011.09.004

mechanism. An optional Collision Avoidance (CA) mechanism is also defined by which a handshake Request to Send (RTS) – Clear to Send (CTS) can be established between source and destination before the actual transmission of data. This CA mechanism aims at reducing the impact of the collisions of data packets and to combat the hidden terminal problem. The DCF can be executed in either ad hoc or infrastructure-based networks and, indeed, is the only access method implemented in most commercial wireless cards. However, the Point Coordination Function (PCF) is also defined in the IEEE 802.11 Standard as an optional polling-based access method for infrastructure-based networks. In PCF, there is no contention to get access to the channel since the access point (AP) polls the stations of the network to transmit data. Therefore, collisions of data packets can be completely avoided. This access method, which may achieve superior performance under heavy traffic loads and can provide some degree of Quality of Service (QoS), has received less attention in the literature.

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

The fact that the DCF of the 802.11 Standard MAC protocol is based on CSMA/CA and BEB as the contention resolution algorithm leads to some underperformance that has been deeply analyzed and improved in the literature over the last years. Some works propose to improve both fairness and throughput by tuning the backoff algorithm at run-time and adjusting it to the load conditions of the network (i.e., the number of active stations and the amount of offered data traffic) [2–4]. These works are inspired by the great impact that the tuning of the size of the contention window has on the performance of the protocol; more precisely, the specific way this size is increased upon collision and decreased upon successful transmission. Other works have focused on adding a power control mechanism to the system, as [5], where the RTS/CTS handshake is transmitted at maximum power while data and ACK are transmitted at the minimum required power, or [6] where, based on the same idea of [5], data senders transmit power spikes during data transmission in order to reduce the probability of having any potential interfering station. Slotting time [7] or using multiple minislot pairs for RTS/CTS handshake [8] are ideas that have been also proposed to improve the performance of the system, among a vast amount of published proposals. Since the fundamentals of the MAC layer have survived across new amendments of the standard, still more progress in the incremental evolution of the IEEE 802.11 MAC protocol is expected to come. However, the simplicity of a CSMA-based protocol comes at the cost of a trialand-error approach where a transmission attempt can result in a collision if several users contend for the access to a common medium as the traffic load of the network increases. For this reason, it seems reasonable to believe that the combination of different access methods, integrating random access for low traffic loads with reservation mechanisms for higher traffic loads, could improve the overall performance of a network. In fact, this approach has been already used in the context of infrastructure-based networks [9–13] combining static Time Division Multiplex Access (TDMA) with dynamic CSMA access. Most of these works deal with different approaches to use empty preallocated TDMA slots by other users with data ready to transmit and which contend for the channel using CSMA. However, up to our knowledge, there are very few works in the literature dealing with this approach in a distributed manner, i.e., for ad hoc networks. This is the main motivation of the work presented in this paper, where we define the Distributed Point Coordination Function (DPCF) as a combination of the DCF and the PCF of the IEEE 802.11 Standard. The goal is to improve the general performance of wireless ad hoc networks under heavy traffic load conditions. We have observed that there are very few works dealing with the PCF, which can indeed potentially achieve better performance than the DCF under heavy traffic conditions. Some contributions related to the PCF improve the overall network performance through novel scheduling algorithms [14–18] or by designing new polling mechanisms that can reduce the overhead associated to the polling process [19]. However, there have been almost no efforts in extending the operation of the PCF to ad hoc

537

networks in order to provide them with some degree of QoS. The only exception can be found in [20] where a virtual infrastructure is created into a MAC protocol called Mobile Point Coordinator MAC (MPC-MAC) in order to achieve QoS delivery and priority access for real time traffic in ad hoc networks maintaining both the PCF and the DCF. In summary, a clustering based mechanism is used to achieve the correct operation of the PCF in a distributed environment. The duration of PCF and DCF periods and the criterion upon which a terminal is chosen to be the MPC (acting as AP) are fixed and they are determined by the MAC protocol configuration. This approach works well in low dynamic environments where the topology does not vary frequently. In this situation the overhead associated to the ‘‘hello’’ messages required for the clustering mechanism can be kept to a minimum. However, it may not be convenient for spontaneous and highly dynamic environments where the clustering overhead could impact negatively on the efficiency of the network. In addition, this protocol does not consider that the responsibility of becoming cluster head should be shared among all the users of the network to ensure certain fairness regarding the extra energy consumption associated to the role of coordinating a cluster. Taking into account this background, we want to contribute to the topic by presenting in this paper the DPCF as an extension of the PCF to operate over infrastructureless networks. By combining the DCF and the PCF using a spontaneous and dynamic clustering mechanism at the MAC layer it is possible to extend the higher performance of the PCF to networks without infrastructure. A preliminary description and performance evaluation of the protocol was already presented in [21]. In this paper, we extend that work in the following way. First, we present a detailed functional description of the protocol as well as a comprehensive performance evaluation based on computer simulation both for single-hop and multi-hop networks. In addition, and in the light of the commercial success of DPCF, we also discuss and quantitatively evaluate the backwards compatibility of DPCF with legacy IEEE 802.11 networks, which was not presented in [21]. It is worth mentioning that the contribution presented in this paper has been inspired by the work presented in [22], where the near-optimum Distributed Queuing Collision Avoidance (DQCA) MAC protocol for WLAN [23] was adapted to be executed in distributed networks without infrastructure. The new resulting protocol was called Distributed Queuing MAC Protocol for Ad hoc Networks (DQMAN). The essential idea of that work is that the DCF protocol is used to create spontaneous clusters. One a node seizes the channel using the rules of the DCF, it becomes master and establishes a temporary cluster where the DQCA protocol can be executed. A similar idea is presented in this paper to extend the operation of the PCF within infrastructure-less networks. The paper is organized as follows. The DCF and the PCF of the IEEE 802.11 Standard are summarized in Section 2. The DPCF protocol is then described in Section 3. In Section 4, we present a comprehensive performance evaluation of the protocol by means of computer simulation both in single-hop and multi-hop settings. The feasibility of

538

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

integrating DPCF stations in a legacy DCF network is evaluated in Section 5. A mixed scenario is presented wherein DPCF and DCF stations can successfully coexist and intercommunicate with each other. Finally, Section 6 concludes the paper and outlines some future lines of research. 2. IEEE 802.11 MAC protocol overview An overview of the operation of the DCF and the PCF of the IEEE 802.11 Standard MAC protocol is included in this section. A comprehensive description of them can be found in [1]. 2.1. DCF overview The DCF is the mandatory coordination function implemented in all standard compliant devices. Two access modes of operation are defined in the DCF: (1) Basic access (BASIC) mode: the station which seizes the channel transmits its data packets without establishing any previous handshake with the intended destination. (2) Collision avoidance access (COLAV) mode: a handshake RTS/CTS is established between source and destination before initiating the actual transmission of data. These RTS and CTS get the form of special control packets. The COLAV access mode

DIFS

is aimed at reducing the impact of collisions of data packets and at combating the presence of hidden terminals. Two examples are illustrated in Figs. 1 and 2 representing the operation of the BASIC and the COLAV access modes, respectively. In summary, any station with data to transmit listens to (senses) the channel for a DCF Inter Frame Space (DIFS). If the channel is sensed idle for this DIFS period, the station seizes the channel and initiates the data transmission (or the RTS transmission in the COLAV mode). Otherwise, if the channel is sensed busy, the station backs off and executes a BEB algorithm by which the size of the contention window is doubled up upon any transmission failure and reset to the minimum value upon success. When a data packet is received without errors, the destination sends back an ACK packet after a Short Inter Frame Space (SIFS). This SIFS is necessary to compensate both propagation delays and radio transceivers turn around times to switch from receiving to transmitting mode. It is worth noting that since a SIFS is shorter than a DIFS, acknowledgments have higher priority than regular data traffic. A relevant feature of the DCF is the Virtual Carrier Sensing (VCS) mechanism. Stations not involved in an ongoing transmission defer from attempting to transmit for the time the channel is expected to be used for an effective transmission between any pair of source and destination stations regardless of the actual physical carrier sensing.

SIFS

DIFS

DATA

Source

CW ACK

Destination Others

CW NAV

Time+ Fig. 1. Example: DCF operation (basic access mode).

DIFS Source

Destination

Others

SIFS

SIFS

RTS

SIFS

DIFS

DATA

CW

CTS

ACK

NAV RTS

CW

NAV CTS NAV DATA

Time+

Fig. 2. Example: DCF operation (collision avoidance mode).

539

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

in a same packet. In PCF, some packets can be combined together in order to reduce the number of MAC and PHY headers and thus increase the efficiency of the communications. In any case, the access to the channel is granted one SIFS after the reception of either the poll or the data packet, respectively. A polled user can either transmit a data packet to the AP or to any other station in the network, establishing a peer-to-peer link. If a polled station has no data to transmit, it responds with a special type of control packet, referred to as NULL packet. An example of PCF operation is illustrated in Fig. 4. In this example, the AP initiates a CFP by transmitting a beacon (B). After a SIFS, it combines a poll packet with data to station 1. Upon the reception of this combined packet, station 1 acknowledges the data packet received and responds to the poll by transmitting a data packet to the AP. Note that this is also a combined packet. Then, the AP acknowledges the data packet received from station 1 and combines a poll packet with data to station 2. Upon the reception of the packet, station 2 acknowledges the packet to the AP and transmits data to station 1. Upon the reception of the packet, station 1 acknowledges the received packet. The CFP is finished with the transmission of a CE packet.

To do so, stations update the Network Allocation Vector (NAV) which accounts for the time the channel is expected to be occupied. This information is retrieved from the duration field attached to the overheard RTS, CTS, and data packets. This mechanism is mainly aimed at combating the presence of hidden terminals.

2.2. PCF overview The PCF can only run on infrastructure-based networks wherein an AP sequentially polls stations to transmit data and thus collisions can be totally avoided. This mechanism was initially designed for the provision of QoS over WLANs. When the PCF is executed, time is divided into Contention Free Periods (CFP), wherein the AP sends poll messages to give transmission opportunities to the stations, and Contention Periods (CP), where the DCF is executed. Since the PCF is an optional coordination function and is not implemented in all standard-compliant devices, DCF periods are necessary to ensure access to DCF stations. The interleaving of CFPs and CPs is illustrated in Fig. 3. As also shown in this figure, a CFP is initiated and maintained by the AP, which periodically transmits a beacon (B). The first beacon after a CP (DCF access) is transmitted after a PCF Inter Frame Space (PIFS). The duration of a PIFS is shorter than a DIFS but longer than a SIFS, providing thus the initialization of a CFP with less priority than the transmission of control packets, but with higher priority than the transmission of data packets. The periodically transmitted beacons contain information regarding the duration of both the CFP and the CP and allow a new arrived station to associate to the AP during a CFP. The CFP is finished whenever the AP transmits a CFP End (CE) control packet. During a CFP, the only station allowed to transmit data is the one being polled by the AP or any destination station which receives a data packet and has to acknowledge (ACK) it, if applicable, and can combine the ACK with data PIFS

The Distributed PCF (DPCF) protocol is presented in this section as an adaptation and extension of the PCF to operate on distributed infrastructureless wireless ad hoc networks. 3.1. Protocol description We consider a set of terminals equipped with WLAN cards forming a spontaneous ad hoc network. Any station must be able to operate in three different modes: idle,

Contention Free Period Polling Access B

Access Point (AP)

3. A new MAC protocol: DPCF

B

Contention Period DCF Access CE

B

NAV

802.11 Station

CFP B

NAV Time +

Fig. 3. IEEE 802.11 PCF interleaves CFPs with CPs.

Contention Free Period (CFP) SIFS Access Point (AP)

Station 1

B

POLL 1+Data (1)

ACK+POLL 2+Data (2)

CE

ACK+Data (AP)

ACK

ACK+Data (1)

Station 2

Time+

Fig. 4. Example: PCF operation.

540

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

Within a cluster, the master can poll the slaves following any arbitrary order. Regardless of the specific polling policy, the master requires some knowledge of the local neighborhood in order to be able to carry out the polling mechanism. To do so, all the stations overhear the ongoing packet transmissions in their vicinity in order to create a neighbor table with an entry for each station in the local neighborhood. This table should be updated along time. The specific scheduling of the polling mechanism is out of the scope of the basic definition of DPCF. Only as an example, a round robin polling scheme can be executed following the entries of the neighbor table. In any case, once a station is polled by the master, it may transmit a data packet to any other slave (peer-to-peer communication model) without routing all the data through the master. Therefore, the master only acts as an indirect coordinator of the communications, but not necessarily as a concentrator of traffic (as the AP does in a regular centralized network). The duration of a cluster is variable and depends on the aggregate traffic load of the network. An inactivity mechanism is considered in DPCF to avoid the transmission of unnecessary polls when there are no more data packets to be transmitted. This mechanism consists of the following: any master maintains a counter that is incremented by one unit upon each NULL packet received from a polled station with no data to transmit. This counter is reset to zero whenever a station responds to a poll with the transmission of a data packet. If the counter gets to a given tunable value, a CE packet is sent and the cluster is broken. On the contrary, it may happen that under heavy traffic conditions once a station becomes master it operates as such for the whole operation of the network due to the absence of idle periods. This would be unfair in terms of sharing the responsibility of being master in the network among all the stations. Therefore, it is necessary to determine an upper-bound for the maximum time that a station can operate as master without interruption. This limit is especially important in infrastructureless networks where fair energy consumption is a must. The approach in DPCF is the following: any master maintains a Master Time Out

master, and slave. Initially, all the stations operate in idle mode but they must be able to change the mode of operation when necessary. Idle stations with data to transmit get access to the channel using the rules of the regular DCF. Whenever a station gets access to the channel, it transmits an RTS targeted to the intended destination of the data packet. This packet initiates a clustering process. Upon the reception of the RTS, the intended destination of the packet becomes master and responds to the RTS with a beacon (B) followed by a poll targeted to the station which transmitted the RTS. A cluster is established and a CFP is initiated inside this cluster. All the idle stations which receive the beacon become slaves and get synchronized to the master. Cluster membership is spontaneous and soft-binding: there are no explicit association and disassociation processes and a station belongs to a cluster as long as it can receive the beacons broadcast by the master. As in the PCF, a cluster is broken when the master transmits a CE packet. Upon the reception of this CE packet, all the slaves revert to idle mode and execute a backoff in order to avoid a certain collision if more than one station has data to transmit. Therefore, according to this operation, the clustering algorithm of DPCF is spontaneous in the sense that the first idle station with data to transmit initiates the clustering algorithm. An example of operation is represented in Fig. 5. In this example, station 1 has data to transmit to station 2. Once the station 1 successfully seizes the channel executing the rules of the DCF, it transmits an RTS to station 2. Upon the reception of the packet, station 2 becomes master and transmits a beacon. The first poll is then sent to station 1, which has a data packet ready to transmit. Station 1 transmits the data packet to station 2. Then, station 2 acknowledges the reception of the packet and polls station N with a combined packet. Since station N has no data packets to transmit, it sends a NULL packet. Finally, station 2 transmits the CE packet to indicate the end of the cluster phase. All the slave stations revert to idle mode and execute a backoff to reduce the probability of collision if more than one station has data to transmit.

Cluster

PCF SIFS

Station 1

Station 2

RTS

DATA for station 2

B

CW ACK + POLL N

POLL 1

CE

NULL

Station N

CW

CW Time +

Station 2 becomes Master, and stations 1,3,…,N become Slave All the stations are idle

Fig. 5. Example: DPCF operation.

All the stations revert to idle mode

541

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

(1) The first RTS triggering the clustering mechanism reserves the channel just for the time required to receive the first beacon from the intended destination (the one that will become master). (2) Each transmitted beacon (broadcast by the master) reserves the channel for the maximum interbeacon time, computed as NpollsMIFS. Recall that the actual interbeacon time is not deterministic and thus the worst case is considered for channel reservation purposes (conservative approach).

(MTO) counter which determines the maximum duration of a cluster. The value of the MTO corresponds to the maximum number of beacons (MTO = Nbeacons) that a master can transmit without interrupting the operation of its cluster. The MTO counter is decremented by one unit after each beacon is transmitted. Whenever the MTO counter expires, a CE packet is transmitted and the cluster is broken regardless of the traffic load or activity of the stations. Therefore, the maximum time that a station can operate as master is denoted by TMAX and can be computed as

T MAX ¼ Nbeacons  Npolls  MIFS ¼ MTO  Npolls  MIFS:

ð1Þ

In order to summarize the whole operation of DPCF, a general flowchart is shown in Fig. 7. The left branch of the chart models the operation of any station becoming master when requested by any other stations and the right branch of the chart represents the operation of the station initiating the clustering algorithm when it has data to transmit.

Npolls denotes the number of polls transmitted between beacons, which can also be tuned, and MIFS is defined as the Maximum Inter Frame Space whose duration corresponds to the maximum possible time between two consecutive polls. The duration of a MIFS can be computed as the time elapsed when: (1) The master station combines an ACK of a recently received data packet with a poll and a data packet. (2) The station polled acknowledges the reception of the data packet from the master and combines the ACK with data for a third station. (3) The third station transmits the ACK of the data packet received from the second station.

4. Performance evaluation In order to evaluate the performance of DPCF, we have implemented the protocol rules in a custom-made C++ link-level simulator [24]. The simulator works in an object-oriented basis and the source code of each station runs in parallel. Each node in the network constitutes a different entity (instance of a class) that executes the code that would be implemented in a real platform. The main motivations for implementing the protocol in a custom-made C++ simulator rather than in any other well known system simulation platform (such as ns-3, for example) are the possibility of isolating the MAC protocol performance from the rest of the network and the faster execution of the simulations as the simulator can be tailored for the specific protocol we want to evaluate. The system parameters have been set according to the one of the modes defined in the PHY layer of the IEEE 802.11 g Standard [1] and they are summarized in Table 1. In the following, we first evaluate the performance of DPCF in single-hop networks, wherein all the stations can sense the same state for the channel (no hidden and exposed terminals), and in multi-hop networks, where not all the stations are in the transmission range of each other.

The definition of a MIFS is illustrated in Fig. 6. Note that it also corresponds to the minimum period of time that a station has to listen to the channel before establishing a new cluster in order to reduce the probability that another master is present. The criteria to set the value of the MTO are highly application-dependant. A greater value of the MTO will lead to greater efficiency in terms of throughput since more time will be used in collision free periods than in contentionbased clustering processes. However, longer clusters may lead to unbalanced energy consumption between those users becoming masters, and those users just behaving as slaves. This tradeoff has to be managed depending on higher-level application requirements. Regarding the channel reservation information (duration fields attached to data and control packets for the execution of the virtual carrier sensing function), the two following simple rules apply to DPCF: MIFS SIFS

Master Station

Slave 1

SIFS

SIFS

ACK+POLL 1 +Data for station 1

POLL 2 ACK +Data for station 2

ACK

Slave 2

Time+ Fig. 6. Definition of MIFS.

542

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

Idle

Idle

Receive RTS from user i

Data to transmit to station i?

no

Set to Master mode yes Transmit First Beacon Binary Exponential BackOff

Transmit RTS to station i

Poll user i

NULL received?

no

yes

Inactivity counter++

Beacon from station i? yes

no Set to Slave mode

Reset Inactivity Counter

Poll next user

MTO counter --

no

yes no

PCF operation

Limit reached?

no CE received?

yes

MTO=0?

yes Transmit CE

Revert to Idle mode

(a) Station becoming master

(b) Station triggering clustering

Fig. 7. DPCF clustering flowchart.

Table 1 System parameters for evaluation of DPCF. Parameter

Value

Parameter

Value

Data packet length (MPDU) Data Tx. rate MAC header SIFS, PIFS, DIFS RTS, BEACON, CF_END and POLL packets CWmin MTO

1500 bytes 54 Mbps 34 bytes 10, 30, 50 ls 20 bytes 16 3

Constant message length Control Tx. rate PHY preamble SlotTime (r), CTS and ACK packets CWmax Polls per beacon

1500 bytes 6 Mbps 96 ls 10 ls 14 bytes 256 19

4.1. Single-hop networks We first consider the case of a single-hop network composed of 20 stations, all of them within the transmission range of each other. All the stations generate data packets of fixed-length following a Poisson arrival distribution and

they contribute equally (homogeneously) to the total aggregate data traffic of the network. The destination of each packet is randomly selected among all the stations of the network with equal probability. In order to focus on the MAC layer, all the packets are assumed to be received without errors and thus the results herein

543

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

In order to further analyze this apparently counterintuitive result, Fig. 9 shows the probability that a station transmits a data packet when it is polled. It has been considered for this calculation that the AP (in the PCF network) and the masters (in the DPCF network) are virtually polled every time they poll a station as they have the possibility to combine the polls with data and ACK packets. The probability of transmitting data when being polled is quite similar in the two networks for low traffic loads. However, this probability is much higher in the DPCF network than in the PCF network for high traffic loads. While the efficiency of the polling in DPCF gets close to 98% for high traffic loads, it remains close to 55% in the PCF network. This efficiency translates directly into a higher efficiency of DPCF, since the ratio of data packets transmitted per control overhead is higher. The reason for these figures is that there is a severe unbalance between the channel access opportunities between the AP and the regular stations in the PCF network. This can be seen in Fig. 10, where we plot again the probability of actually transmitting when being polled. Now, two different curves for the PCF network are represented corresponding to the average probability among of all the regular stations and to the probability for the AP alone, separately. The AP has a channel access opportunity every time it polls another station, but most of these transmission opportunities are not used for the actual transmission of data (note that the probability of transmitting when being polled is below 10% in all cases for the AP), decreasing the overall efficiency of the polling mechanism. This unbalance between the AP and the stations is avoided in DPCF by sharing dynamically and spontaneously in time the responsibility of being master among all the stations of the network. It is well known that the DCF is fair in the long-term, and so is the clustering algorithm of DPCF. Since all the stations of the network get the role of master periodically, the unbalanced access of the AP in the PCF network is shared in the DPCF network. Every time a station is set to master it can transmit all its backlogged data packets and thus take advantage of the

presented correspond to an upper-bound of the performance of the protocol (from the wireless channel effects point of view). It is also assumed that an ideal round robin scheduling is performed to poll all the stations once a cluster is established. Three different networks have been studied (they all have been implemented in the simulator): (1) DCF: a network wherein all the stations only execute the DCF with the collision avoidance access method. (2) PCF: a network wherein an AP manages the access to the channel. However, stations transmit directly to the intended destination without routing traffic through the AP. In this network, we consider that the AP also has data to transmit as any other regular station. (3) DPCF: a network wherein all the stations execute the proposed DPCF protocol. According to the parameters presented in Table 1, the number of polls between beacons has been set to 19 and it indicates that all the slaves within a cluster are polled exactly once by the master between the transmission of two consecutive beacons. In addition, the setting MTO = 3 indicates that all the slaves are polled at most three times when a cluster is established unless the inactivity mechanism is triggered by the master. The throughput of the three different networks is plotted in Fig. 8 as a function of the total aggregate offered load to the network. As expected, the three curves grow linearly until they reach the saturation throughput. The three protocols are stable for heavy traffic conditions without entering in congestion and thus they can operate under sporadic situations of high peak traffic loads without collapsing the network. The saturation throughput of DPCF is remarkably higher than that of DCF, achieving an improvement of approximately 250%. Collisions and backoff periods are reduced in the DPCF network compared to the DCF network, thus yielding higher performance. In addition, the performance of DPCF is even superior to the regular PCF, attaining 25% higher saturation throughput. 30

DPCF PCF DCF

27 24

Throughput (Mbps)

25 % 21 18

250 %

15 12 9 6 3 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

Total Offered Traffic Load (Mbps) Fig. 8. Throughput comparison DPCF, PCF, and DCF in a single-hop network.

32

34

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

Probability of transmitting when being polled

544

1 DPCF PCF

0.9 0.8

45 %

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

32

34

Total Offered Traffic Load (Mbps)

Probability of transmitting when being polled

Fig. 9. Probability of transmitting when being polled in a single-hop network.

1,1 Stations in PCF network Stations in DPCF network AP in PCF network

1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

Total Offered Traffic Load (Mbps) Fig. 10. Probability of transmitting when being polled in a single-hop network.

prioritized access to empty its data buffers while operating as master. Indeed, the fact that a station operating in master mode has more channel access opportunities than a slave station can be seen as an implicit mechanism to provide with some incentive to stations to become master despite the extra actions they must carry out and the corresponding increase in energy consumption. The performance in terms of average packet transmission delay is plotted in Fig. 11. We define this delay as the average time elapsed since a packet arrives at the MAC layer until it is successfully acknowledged by the intended destination. It is worth seeing that for low offered loads, the best performance is attained in the DCF network. This is an expected result since every time a station has data to transmit it can successfully seize the channel immediately without needing to wait for being polled (the probability of finding the channel busy and the probability of collision are low due to the low offered traf-

fic load). However, as the offered load grows, the average packet transmission delay in the DCF network grows sharply for traffic loads above 10 Mbps. On the other hand, the DPCF attains average delays below 200 ms for traffic loads up to 22 Mbps, increasing the throughput of the standard DCF network and attaining superior performance than the PCF. These results confirm the idea that PCF-like mechanisms are worthy when the traffic load and the number of transmitting stations are relatively high. 4.2. Multi-hop networks Although the DPCF protocol has been designed with the motivation of improving the performance in highly-loaded single-hop networks, it can also operate in multi-hop settings. In this section we evaluate a simple multi-hop case study to show that the protocol can indeed operate in networks where not all the nodes are in the transmission

545

Average Packet Transmission Delay (ms)

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

200 180 160 140 120 100 80 60 40

DCF PCF DPCF

20 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

Total Offered Traffic Load (Mbps) Fig. 11. Average packet transmission delay in a single-hop network.

range of each other. Therefore, since the protocol is based on the DCF of the 802.11 Standard, effects such as the presence of hidden or exposed terminals may appear. We have simulated a tandem network formed by 5 static stations set in line and equally spaced as the one represented in Fig. 12. The distance between the stations, the transmission powers, and the channel propagation parameters have been adjusted so that:

A collision occurs if two simultaneous transmissions are received within either the transmission or the interference range of the transmitters. We assume that all the stations have perfect routing information and thus route the packets through the station in its transmission range that is closer to the intended destination. The rest of the parameters have been set as in the previous section for the single-hop evaluation. The total throughput delivered to destination is plotted in Fig. 13 as a function of the total offered load to the network, assuming homogeneous distribution of the load among the nodes of the network. Note that the traffic delivered to the intermediate stations in a multi-hop route is not accounted for this calculation. The curves show that DPCF outperforms DCF for all traffic loads. Indeed, for low traffic loads both protocols behave almost identically

(1) Every station can transmit directly to immediate neighbors at one-hop distance. (2) Every station at two hops of a transmitting station can sense the channel busy, but cannot decode the transmitted information. (3) Every station at three hops of a transmitting station is oblivious to the transmission.

Transmission Range IDLE

IDLE

IDLE

SLAVE

Interference Range

Fig. 12. Tandem multi-hop network.

MASTER

546

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

extensive computer simulation the performance of a compound scenario where DPCF stations coexist and intercommunicate with legacy DCF stations.

delivering all the data traffic offered to the network. However, while DCF saturates around 5 Mbps, DPCF is capable of delivering up to 9 Mbps (80% higher saturation throughput), almost doubling up the capacity of the legacy DCF. Comparing these results to the ones obtained for the single-hop case, it is possible to see that the total offered load that can be conveyed in the multi-hop network is considerably lower. This is mainly due to the fact that in the multihop environment some packets need to travel along several hops to get to the final destination. The average packet transmission delay is plotted in Fig. 14 for both the DPCF and the DCF networks. In this case, this measure is defined as the average time elapsed from the moment a packet arrives at the MAC layer of the source station until it is successfully delivered to the final destination (end-to-end time). The curves show that the DCF attains lower average packet transmission delay for low traffic loads. Two are the main reasons for this lower average delay. First, the longer MIFS of DPCF (compared to the DIFS of DCF) adds latency to all the transmissions, increasing the average packet transmission delay in the DPCF network for low traffic loads. In addition, in the DPCF network, slaves cannot transmit immediately whenever they have data to transmit but they have to wait to be polled by a master, increasing thus the average access delay. However, note that the average delay is lower than 300 ms for loads up to 8 Mbps in the DPCF network and it gets unbounded in the DCF network for traffic loads over 6 Mbps. Therefore, the DPCF protocol attains better performance when the traffic load of the network is higher, attaining up to 25% better performance than the DCF in this multi-hop setting.

5.1. Coexistence methodology We consider a mixed network formed by two different kinds of stations, namely, DPCF and DCF stations. We assume that DPCF stations can execute the rules of DPCF as well as the rules of the DCF. In fact, the rules of the DCF are included in DPCF. For this reason, this kind of stations will be hereafter referred to as either DPCF or dual stations. On the other hand, DCF stations can only execute the rules of the DCF and are not aware of the presence of DPCF stations and, indeed, they do not know the rules of DPCF. This latter group represents already deployed equipment which has been shipped without knowledge of the existence of DPCF. By default, all the stations execute the DCF to get access to the channel. However, a DPCF station which gets the channel can either (i) transmit data following the rules of the DCF or (ii) initiate a cluster following the rules of DPCF described in Section 0. Therefore, it is necessary to include a stamp in the RTS packets to indicate whether the transmitting station wills to initiate a cluster or not. To do so, it is possible to use either bit B8 or B9 of the Frame Control (FC) field of RTS packets, which are only used in infrastructure-based communications [1]. One of these bits will be used as a flag to distinguish between regular RTS packets and what we will call dual-RTS packets, which are associated to the initialization of a DPCF clustering phase. A key advantage of this approach is that DCF stations will ignore this field and will thus decode a dual-RTS packet as a regular RTS packet. If a cluster is established, all the DCF stations must update their NAV with the value of the duration field attached to the beacons transmitted by the master and wait for the transmission of the CE packet indicating the end of the cluster. However, neither beacons nor CE packets can be decoded by DCF stations. In order to overcome this

5. Coexistence with legacy IEEE 802.11 networks For the commercial success of any new technology it is necessary to ensure a certain degree of backwards compatibility. For this reason, we present in this section a simple methodology to enable the coexistence of DPCF with legacy DCF networks. In addition, we evaluate through 10 DPCF

Throughput to destination (Mbps)

9

DCF

8 7 80% 6 5 4 3 2 1 0 1

2

3

4

5

6

7

8

9

10

Total Offered Traffic Load (Mbps) Fig. 13. Throughput to destination of DPCF in a multi-hop network.

11

12

547

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

Average Packet Transmission Delay (ms)

300 270 240 210 180 150 120 90 60 DPCF

30

DCF

0 1

2

3

4

5

6

7

8

9

10

11

12

Total Offered Traffic Load (Mbps) Fig. 14. Average packet transmission delay of DPCF in a multi-hop network.

(1). It is worth mentioning that these forced DCF periods can be also used by DPCF stations to overhear ongoing transmissions and learn the composition of the network in order to optimize the polling mechanism whenever a cluster is established. The interleaving of clustering and DCF periods is illustrated in Fig. 15. In the next section, we evaluate through computerbased simulation the performance of a mixed scenario composed by DPCF and legacy DCF stations to show that the mechanism works and DPCF can coexist with already deployed WLANs.

problem, both beacons and CE packets can be substituted by what we will call dual-CTS packets, defined as regular CTS packets with the following two modifications: (1) The duration field of the CTS has the value of the interbeacon interval if it is used as a beacon and the value zero if it is used as a CE packet. (2) As in the case of dual-RTS packets, a flag bit is included to distinguish CTS packets acting as beacons from regular CTS packets. In addition, it is necessary to ensure that DPCF stations do not capture the channel. Note that DCF stations will be unable to get access to the channel if DPCF stations establish a cluster every time they seize it. Therefore, DPCF stations must not be allowed to initiate a cluster before a certain time has elapsed from the transmission of the last CE packet, i.e., the end of the last cluster. To this end, we define the Equivalent Cluster Time (ECT) as the minimum DCF operation time between two consecutive clusters. For the sake of simplicity, this time is measured in MTO units and can be easily translated into time duration using

5.2. Performance evaluation We have used the computer-based simulator to evaluate the performance of a network formed by a total of 20 stations in a single-hop layout, all of them in the transmission range of each other. 5 out of these 20 stations are DPCF stations while the other 15 stations are legacy DCF stations. The rest of the configuration parameters have been set as for the single-hop performance evaluation of DPCF presented before but with the following modifications: DPCF stations

DCF operation during a ECT DIFS

DPCF Station

Cluster

CW

DCF operation

Cluster

DPCF Station

Cluster

CW

DCF operation

Cluster

DCF Station

NAV

CW

DCF operation

NAV

DCF Station

NAV

CW

DCF operation

NAV

Time+ Fig. 15. DPCF and DCF periods are interleaved.

548

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

28 DPCF Mixed network ECT=1 DCF

Througput (Mbps)

26 24 22 20 18 16 14 12 10 8 6 4 2 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

Total Offered Traffic Load (Mbps) Fig. 16. Throughput in a mixed scenario DPCF-DCF.

14 DCF stations, DCF operation DCF stations, ECT=3 DCF stations, ECT=2 DCF stations, ECT=1

13 12

Throughput (Mbps)

11 10 9 8 7 6 5 4 3 2 1 0 0

3

6

9

12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60

Total Offered Traffic Load (Mbps) Fig. 17. Throughput of DCF stations in a mixed scenario.

transmit 16 poll packets in between two consecutive beacons and the MTO has been set to 4. In order to obtain the results presented in this section, the mixed scenario as well as two benchmark networks where the 20 stations execute only the DPCF or the DCF, respectively, have been simulated. The throughput as a function of the total offered traffic load is illustrated in Fig. 16 for the three considered networks. The ECT has been set to 1, i.e., the duration of the forced DCF operation period after a DPCF cluster is equal to the maximum duration of the cluster time. As shown in the figure, the performance of the mixed network lies in between the performance of the DCF and the DPCF networks. Since both protocols are alternately executed in a mixed network, it seems reasonable that the results show an intermediate performance. Therefore, the main conclusion is that the mixed network works properly and, as a consequence, the coexistence and a smooth migration to the new protocol are actually feasible.

It is interesting to analyze the throughput performance perceived separately by the group of DPCF and DCF stations. To do so, the throughput of the group of 15 DCF stations is illustrated in Fig. 17, while the performance of the group of 5 DPCF stations is illustrated in Fig. 18. As a benchmark performance, the curve labeled DCF operation corresponds to the situation where only the DCF is executed. Note that the values of the x-axis in these curves correspond to the total offered traffic load considering the contribution of all the stations regardless of whether they are DCF or DPCF stations. It is worth seeing that the performances perceived by each group of stations are complementary. As the value of the ECT increases, i.e., longer DCF periods are allowed, the performance of the DCF stations increases, as expected. In the limit, where the DCF tends to be the only access method used by all the stations (including the dual DPCF stations), the performance of the DCF group of stations is exactly 3/4 of the total throughput of the network (which

549

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

14 DPCF stations, ECT=1 DPCF stations, ECT=2 DPCF stations, ECT=3 DPCF stations, DCF operation

Throughput (Mbps)

13 12 11 10 9 8 7 6 5 4 3 2 1 0 0

3

6

9

12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60

Total Offered Traffic Load (Mbps) Fig. 18. Throughput of DPCF stations in a mixed scenario.

Average Packet Transmission Delay (ms)

150 140 130 120 110 100 90 80 70 60 50 40 30 Mixed network ECT=1 DCF DPCF

20 10 0 0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

Total Offered Traffic Load (Mbps) Fig. 19. Average packet transmission delay in a mixed scenario.

corresponds to 15 out of 20 stations). This is due to the well known long-term fairness of the DCF. On the other hand, the throughput perceived by the group of DPCF stations is higher as the value of the ECT becomes smaller. A smaller value of the ECT allows the occurrence of more frequent DPCF periods where the performance is higher than the one within DCF periods. It is worth noting that the performance of the DPCF stations can increase up to 400% when the ECT = 1 with respect to the case of only executing the DCF. Finally, the average packet transmission delay of the network as a function of the total offered load to the network is illustrated in Fig. 19 (with no distinction between DPCF and DCF stations, and considering the overall average). For low traffic loads, the delay of the mixed scenario is similar to that of the DCF network. However, as the traffic load grows, the delay of the mixed scenario grows sharply. This is due to the fact that DCF stations suffer from long access delays when a cluster is established frequently

and they have to wait for the DCF period to get access to the channel. According to these results, the value of the ECT plays a key role in the performance perceived by each of the group of stations, and thus the fairness between the DCF and the DPCF stations. The higher the ECT, the fairer the network is. However, the lower the overall network performance. This tradeoff should be handled differently depending on the application or the priority that DPCF users should have. In some applications it might be interesting to give higher priority to the DPCF stations, while in other situations it will be more convenient to ensure fairness, letting the DPCF operation to run only when there are no legacy stations.

6. Conclusions DPCF has been presented in this paper as the extension of the PCF of the IEEE 802.11 Standard to operate in distrib-

550

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551

uted wireless ad hoc networks without infrastructure. The main idea of DPCF is that the stations of the network get access to the channel by executing the rules of the DCF. Any station which seizes the channel transmits its data and also establishes a temporary dynamic cluster to manage the pending transmissions of all the neighbors with data ready to be transmitted. The key of this mechanism is that there is no cluster head selection, but clusters are created in a spontaneous manner. This reduces the control overhead to establish a fixed clustering architecture and increases the capability of the network to dynamically adapt to the unpredictable nature of ad hoc networks. Comprehensive performance evaluation of the protocol through link-level computer simulation shows that the new proposal improves the performance of ad hoc networks when compared to current standards. In addition, we have evaluated in this paper the backwards compatibility of DPCF with legacy IEEE 802.11 networks. We have presented a simple methodology that facilitates the coexistence of DPCF with legacy 802.11 networks. A performance evaluation of a mixed network wherein DPCF stations coexist with legacy DCF stations shows that the two protocols can feasibly coexist. By ensuring minimum periods of time wherein all the stations only execute the DCF access, fairness can be attained. Indeed, the results show that by properly balancing the maximum time allowed for clusters and the minimum time reserved for forced DCF access, the performance of the overall network can be tuned at will. Actually, there is a tradeoff between fairness among the different types of stations (DCF and DPCF) and the overall network performance. This trade off might be driven by the specific requirement of each network. Just as a simple example, some preselected stations could execute DPCF only for the transmission of critical control information, while the DCF can be used for the regular transmission of data packets without specific requirements of QoS. The results presented in this paper are rather promising and, in fact, future work will be aimed at optimizing the design of the DPCF and at implementing the protocol in a testbed to evaluate its actual performance in a real environment.

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11] [12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

Acknowledgements This work has been partially supported by the research projects EXALTED (ICT-5-258512), CENTENO (TEC200806817-C02-02), CO2GREEN (TEC2010-20823) and C2POWER (ICT-248577).

References [1] IEEE, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Std. 802.11 - 2007. [2] G. Bianchi, L. Fratta, M. Oliveri, Performance Evaluation and Enhancement of the CSMA/CA MAC protocol for 802.11 Wireless

[21]

[22]

[23]

[24]

LAN’s, in: Proc. of the IEEE PIMRC 1996, Taipei, Taiwan, October 1996, pp. 392–296. J. Weinmiller, M. Schläger, A. Festag, A. Wolisz, Performance study of access control in wireless LANs-IEEE 802.11 DFWMAC and ETSI RES 10 HIPERLAN, Mobile Networks Appl. 2 (1997) 55–67. F. Cali, M. Conti, E. Gregori, Dynamic Tuning of the IEEE 802.11 Protocol to Achieve a Theoretical Throughput Limit, IEEE/ACM Trans. Network. 8 (6) (2006) 785–799. S. Agarwal, S. Krishnamurthy, R.H. Katz, S.D. Kao, Distributed power control in ad hoc wireless networks, in: Proc. of the IEEE PIMRC, 2001. E.S. Jung, N.H. Vaydia, A Power Control MAC protocol for ad hoc networks, in: Proc. of the ACM Mobicom, Atlanta, GA, September 2002, pp. 36–47. Z. Tang, J.J. Garcia-Luna-Aceves, A protocol for topology-dependant transmission scheduling in wireless networks, in: Proc. of the IEEE WCNC, vol. 3, New Orleans, LA, September 1999, pp. 1333–1337. S. Toumpis, A. Goldsmith, Performance, optimization, and crosslayer design of media access control protocols for wireless ad hoc networks, in: Proc. of the IEEE ICC, Anchorage, AK, May 2003, pp. 2234–2240. D.J. Goodman, R.A. Valenzuela, K.T. Gayliard, B. Ramamurthi, Packet reservation multiple access for local wireless networks, IEEE Trans. Commun. 37 (8) (1989). I. Chlamtac, A. Faragó, A.D. Myers, V.R. Syrotiuk, G.V. Záruba, ADAPT: A dynamically self-adjusting media access control protocol for ad hoc networks, in: Proc. of the GLOBECOM, December, 1999, pp. 1115. A. Rhee, M. Warrier, J. Min, ZMAC: A hybrid MAC for wireless sensor networks, in: Proc. of Sensys, San Diego, California, 2005. M. Shakir, I. Ahmed, P. Mugen, W. Wang, Cluster Organization based Design of Hybrid MAC Protocol in Wireless Sensor Networks, in: Proc. of the Third International Conference on Networking and Services, June 2007, pp. 78–83. A. Muir, J.J. Garcia-Luna-Aceves, An efficient packet sensing MAC protocol for wireless networks, Springer Mobile Networks Appl. 3 (2) (1998) 221–234. A. Kanjanavapastit, B. Landfeldt, A performance investigation of the modified PCF under hidden station problem, in: Proc. of the ICCCAS 2004, vol. 1, June 2004, pp. 428–432. B. Anjum, S. Mushtaq, A. Hussain, Multiple Poll Scheme for Improved QoS Using IEEE 802.11 PCF, in: Proc. of the IEEE INMIC’05, December 2005, pp. 1–6. D. Ping, J. Holliday, A. Celik, Dynamic scheduling of PCF traffic in an unstable wireless LAN, in: proc. of the CCNC. 2005, January 2005, pp. 445–450. K. Byung-Seo, K. Sung Won, W. Yuguang Fang, Two-step multipolling MAC protocol for wireless LANs, IEEE J. Sel. Areas Commun. 23 (6) (2005) 1276–1286. K. Young-Jae and S. Young-Joo, Adaptive polling MAC schemes for IEEE 802.11 wireless LANs, in: Proc. of the VTC 2003, vol. 4, April 2003, pp. 2528–2532. A. Kanjanavapastit, B. Landfeldt, An analysis of a modified point coordination function in IEEE 802.11, in: Proc. of the IEEE PIMRC’03, vol. 2, 2003, pp. 1732–1736. Y. Tiantong, H. Hassanein, H.T. Mouftah, Infrastructure-based MAC in wireless mobile ad-hoc networks, in: Proc. of the 27th Annual IEEE Conference on Local Computer Networks, 2002, pp. 821–830. C. Crespo, J. Alonso-Zárate, L. Alonso, Ch. Verikoukis, Distributed Point Coordination Function for Wireless Ad hoc Networks, in: proc. of the VTC Spring 2009, Barcelona, Spain. J. Alonso-Zárate, E. Kartsakli, L. Alonso, Ch. Verikoukis, Performance Analysis of a Cluster-Based MAC Protocol for Wireless Ad Hoc Networks, EURASIP Journal on Wireless Communications and Networking, Special Issue on Theoretical and Algorithmic Foundations of Wireless Ad Hoc and Sensor Networks, vol. 2010, Article ID 625619, 16 pp. March 2010. J. Alonso-Zárate, E. Kartsakli, A. Cateura, C. Verikoukis, L. Alonso, A near-optimum cross-layered distributed queuing protocol for wireless LAN, IEEE Wireless Commun. Mag. Spec Issue MAC Protoc. WLAN 15 (1) (2008) 48–55. The MAC Simulator for Wireless Networks (MACSWIN). .

J. Alonso-Zárate et al. / Ad Hoc Networks 10 (2012) 536–551 Jesús Alonso Zárate received his M.Sc. and Ph.D. degrees in Telecommunication Engineering from the Universitat Politècnica de Catalunya (UPC, Spain) in March 2004 and February 2009, respectively. From 2004 to 2005, he worked as an Information Technology consultant at Everis (former DMR Consulting). In 2005, he was granted by the Centre Tecnològic de Telecomunicacions de Catalunya (CTTC) to obtain the Ph.D. on Signal Theory and Communications at the UPC. He is now with the CTTC holding a Research Associate position within the Intelligent Energy Area. He has published several scientific papers in renowned international journals and international conferences over the last years and he has published several book chapters in books devoted to wireless communications. He actively participates in both competitive public funded and industrial research projects. He is currently leading the research project CO2GREEN funded by the Spanish Government. CO2GREEN is focused on the study of cooperative and cognitive techniques to improve the energy efficiency of wireless communications. He is member of the IEEE ComSoc CSIM Technical Committee (Communication Systems Integration and Modelling) and works as reviewer and chair for numerous international conferences. He is part of the Editorial Board of the IET Wireless Sensor Systems Journal and supervises a number of undergraduate and Phd Students. Lately, he has been giving various invited talks in conferences related to Machineto-Machine (M2M) communications. In June 2011, he received the ‘‘Best Ph.D. Thesis Award of UPC’’ in the field of Telecommunications for his thesis read during the course 2008-2009. In addition, also in June 2011, he received the Best Paper Award at the IEEE International Conference on Communications (ICC) 2011 with a contribution towards the energyefficiency of wireless communications.

Cristian Crespo Cintas received his M.Sc. degree in Telecommunications engineering, and Master of Science in Telecommunications Engineering and Management (MASTEAM), from the Universitat Politècnica de Catalunya (UPC, Spain), in July 2010. From 2008 to 2009, he collaborated with the Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), where he developed his Bachelor’s degree related with the Implementation of new Medium Access Control Protocols for Wireless Ad-Hoc networks, and he published a conference paper in the IEEE VTC in April 2009. In 2010, he was involved in the Erasmus programme at Aalborg University, Denmark, where he developed his final Master thesis based on probabilistic location methods for hybrid UWB-RFID systems. Since November 2010, he is working as Specialist of Design and Optimization for Vodafone Spain.

Charalabos Skianis is currently Assistant Professor in the Department of Information and Communication Systems at the University of the Aegean in Samos, Greece. He holds a Ph.D. degree in Computer Science, University of Bradford, United Kingdom and a B.Sc. in Physics, Department of Physics, University of Patras, Greece. His current research activities take upon Novel Internet Architectures and Services, Cloud Computing & Networking, Energy & Context aware Next Generation Networks and Services, management aspects of mobile and wireless networks, ubiquitous and pervasive computing and End-to-End Quality of Service provisioning in heterogeneous networks environment. He has been actively working on the area of computer and communication systems performance modeling and evaluation where he has introduced alternative methodologies for the approximate analysis of certain arbitrary queuing network models. He is also keen in traffic modeling and characterization, queuing theory and traffic control of wired and wireless telecommunication systems. His work is published in journals, conference proceedings and as book chapters and has also

551

been presented in numerous conferences and workshops. He acts within Technical Program and Organizing Committees for numerous conferences and workshops (e.g., IFIP Networking, symposium chair for several IEEE Globecom and IEEE ICC conferences) and as a Guest Editor for scientific journals (e.g., IEEE Networks magazine). He is at the editorial board of journals (e.g., IEEE Wireless Communications, IEEE Communications Magazine, IEEE Surveys & Tutorials, Ad hoc Networks), a member of pronounced professional societies (senior member of IEEE) and an active reviewer for several scientific journals. He is an active member of several Technical Committees within the IEEE ComSoc (e.g., chair for the CSIM technical committee, vice chair for Information Infrastructure technical committee), member of the education board of IEEE ComSoc, and member of IEEE BTS; IEEE TVT and IEEE CS. He has acted as Technical Manager for ICT FP7 VITAL++ project and representative for the University on ICT FP7 HURRICANE project. He is currently project coordinator for ICT FP7 PASSIVE project.

Luis Alonso developed his Ph.D. thesis in the Department of Signal Theory and Communications of the UPC and he reached the Ph.D. degree in 2001. In 2006, he obtained a permanent tenured position in the University, becoming an Associate Professor within the Radio Communications Research Group. In 2009, he has been co-founder of the Wireless Communications and Technologies Research Group (WiComTec), to which currently belongs. He has participated in several research programs, networks of excellence, COST actions and integrated projects funded by the European Union and the Spanish Government, always working on the design and analysis of different mechanisms and techniques to improve wireless communications systems. He has also collaborated with some telecommunications companies as Telefónica, Alcatel and Sener, working as a consultant for several research projects. He has been the Project Coordinator of two Marie Curie Actions (ToK and IEF), funded by the European Union) and he is currently the Coordinator of two more Marie Curie Actions (IAPP and ITN). He has been the Scientist in Charge of one three-year Research Project funded by the Spanish Ministry of Science and Technology, and he is currently the Scientist in Charge of another project that is being carried out in coordination with ADIF, which is the Spanish Railway Infrastructure Administrator. This project is related to wireless communications in the high-speed railway transportation system. He is author of nearly thirty papers in international journals and magazines, one book, five chapters of books, and more than sixty papers in international congresses and symposiums. His current research interests are still within the field of medium access protocols, radio resource management, cross-layer optimization, cooperative transmissions, cognitive radio and QoS features for all kind of wireless communications systems.

Christos Verikoukis got his degree in Physics and his M.Sc. in Telecommunications Engineering in the Aristotle University of Thessaloniki (AUTh) in 1994 and 1997, respectively. He got his Ph.D. from the Signal Theory and Communications Department of the Technical University of Catalonia (UPC), Barcelona. From March 2003 to February 2004 he was research associate of the Southeastern Europe Telecommunications & Informatics Research Institute (INA), Thessaloniki, Greece. Since February 2004 he is a senior researcher in the CTTC. Dr Verikoukis has participated or is currently participating in several European funded projects while he has served as the technical manager in five Marie Curie/PEOPLE and in two Celtic projects. He has also served as the principal investigator of national projects in Greece and in Spain. He has published more than 30 journal papers, 2 books, 11 chapters in different books and 2 patents. His research interests include energy efficient wireless systems, short range communications, cooperative relaying and body area networks.