Complementing 3G Cellular Networks by Multi Hop Capabilities of an Enhanced MAC Protocol for Wireless LAN Frank H.P. Fitzek† , Federico Bertocchi ‡ , Michele Zorzi‡ †
‡
Department of Communications Technology, Aalborg University Neils Jernes Vej 12, 9220 Aalborg Øst, Denmark, e-mail:
[email protected] University of Ferrara, Via Saragat 1, I-44100 Ferrara, Italy, e-mail: [fbertocchi|zorzi]@ing.unife.it
Abstract— Interworking between wireless LANs and 3G networks is expected to break new ground for sophisticated business models. The combination of the 3G core network infrastructure (with service centers and accounting capabilities) and the achievable data rate of WLANs give the possibility to offer new services. WLAN hotspots are on the way, but the high installation costs of WLAN, especially for larger hotspots, repress a faster evolution. The installation costs are mainly driven by the cabling infrastructure. In the past we have already introduced an enhanced MAC protocol for multi hop networks without any infrastructure. In this paper we investigate the MAC performance for a multi hop scenario, enabling low costs WLAN networks with infrastructure that are coupled with 3G networks. This approach should allow network providers to have a fast return on their investments. As video services are expected to be one of the future key–services, we present a detailed performance evaluation of different network topologies transmitting high quality video traffic.
technology such as WLAN at high costs. Even if the services are provided by the 3G core network, the cost of the WLAN infrastructure (cabling, building license, etc) is still high. It is estimated that for very large WLAN hotspots the cabling covers up to 75% of the whole network costs. This percentage becomes even larger in the future if the Access Points (AP) becomes even cheaper.
I. I NTRODUCTION Both 3G and WLAN networks are installed at the moment all over the world. Because of their disjunctive disadvantages the interworking between wireless LANs and 3G networks is expected to break new ground for sophisticated business models. While 3G systems (in conjunction with existing 2 and 2.5G networks) offer large coverage and a rich network infrastructure (application, billing, mobility), WLAN has its potential in high data rates allowing fast web surfing, high quality video transmission and gaming applications. Even the 3rd Generation Partnership Project (3GPP) has identified this interworking as vital for 3G networks introducing [1]. The question arises why WLAN without 3GPP interworking has this limited success for public access networks. The vicious circle can be described as follows: i.) WLAN networks have high installation costs ii.) not many customers are willing to pay for WLAN networks without any real services (email and web are the only one at the moment) iii.) network providers are poor-spirited to invest in such a business case. To break this circle up we have to bring services to the WLAN and reduce the installation costs. While the former task is done by the coupling with 3G networks, the latter is realized by multi–hop networks. Not yet solved are the high installation costs of WLAN. After spending a lot of money onto the 3G spectrum the potential network operators are not willing to set up a parallel
Fig. 1.
Interworking between multi hop WLAN systems and 3G network.
Therefore we advocate the usage of multi hop WLAN systems. In Figure 1 (in accordance with [1]) we present a possible interworking between multi hop WLAN systems and 3G networks. The 3G networks such as Universal Mobile Telecommunications System (UMTS) have their own wireless access such as the UMTS Terrestrial Access Network (UTRAN) and for larger coverage extension the Enhanced Radio Access Network (ERAN). Both networks are connected to the UMTS core network via the Serving GPRS Serving Node (SGSN). The UMTS core supports several services such as billing and charging functionality, mobility, and some sorts of application servers (a prominent representative is the multimedia broadcast multi-cast service (MBMS)). The access performance is rather weak in terms of bandwidth and delay compared to WLAN. But as WLAN does not have any service
infrastructure, we advocate the combination of 3G and WLAN networks. Different business models for 3G/WLAN interworking are reasonable. One possibility is that network providers that have already 3G licenses start also to deploy WLAN networks. Some global players have already started it. At the moment they target only the more interesting places such as airports and hotels. Smaller providers have their own WLAN networks. The realization in terms of coverage, data rate, and security differ among the network providers. The WLAN network can be coupled to the 3G network over an entity called Packet Data Gateway (PDG). It is still under discussion if this entity will be part of the Gateway GPRS Serving Node (GGSN) or a stand alone one. This discussion is out of the scope of this paper, but we had to address this functionality as it is a vital one for the interworking. Different WLAN networks can be connected to the PDG gaining from the service infrastructure of the UMTS core network. As stated in the beginning our main focus in this paper is to have low cost WLAN network solutions. An enabler for low cost WLANs is to use multi–hop capabilities. As given in Figure 1 we introduce some Virtual Access Points (VAP). These are multi–hop enabled terminals that belong to the network providers. These entities have fixed power supply (no energy problem) and no mobility (routing becomes easy). Nevertheless we want to stress that also the multi–hop over mobile terminals should be possible. It is out of the scope of this paper to discuss the impacts of such an approach. The network provider can start with only one fixed access points and several VAPs. If the traffic in the WLAN is increasing (and therefore the number of customers), some of the VAPs can be connected in a wired fashion (becoming a normal AP, some investment has to be done but this should be justified by the higher number of users).
we introduced the random approach. Here N multiple channels are used randomly. Therefore higher bandwidth can be achieved, but the node density decreases as one terminal can send or receive only on one channel at the same time. Our performance evaluation done so far focuses on ad hoc networks. Multiple senders convey data to randomly chosen receivers. This may lead to areas in the network topology that are highly loaded. Throughout the paper we refer to this scenario as Scenario A (referring to ad hoc). For the interworking with 3G networks our main focus in this paper relies on multi hop enabled networks with a given infrastructure. Three different topologies are under investigation, namely: i.) multiple senders transmit (with the help of multi hop enabled terminals or VAPs) to one dedicated receiver (access point at location X in Figure 1). We refer to this as Scenario 1R, ii.) one sender (access point at location Y in Figure 1) transmits to multiple receivers. We refer to this as Scenario 1S, and iii.) multiple senders (access points at location Z in Figure 1) transmit to multiple receivers. We refer to this as Scenario 5S. We lay more emphasis on the down– link case as we believe that this is more business driven. As one prominent candidate for new services in 3G and WLAN, we investigate the MAC performance of different schemes for streaming video. A lot of research has be done in the field of video communication over ad hoc networks or the interworking between ad hoc and cellular systems. In [4], [5] a five phase reservation protocol was introduced using a single channel for ad hoc communication (video services were not addressed). The work of [6], [7], [8] was focusing on video services over ad hoc networks, while [9] and others highlight the interworking of cellular systems and ad hoc. In this paper we try to combine these three research topics and motivate the usage of a new MAC scheme enabling efficient multi hopping.
II. S YSTEM UNDER I NVESTIGATION
III. M ETHODOLOGY AND M ETRIC
In our former work [2], [3] we have presented a new MAC protocol for multi hop enabled WLAN networks based on IEEE802.11 technology. Our new scheme is based on the usage of multiple IEEE802.11 channels using a four way handshake over a common signalling channel, while data transmission occurs on N multiple dedicated channels. In the following we refer to such a channel assignment as dynamic approach. The four way handshake contains two Ready To Send (RTS) packets, two Clear To Send (CTS) packets, and one probing packet. The purpose of the handshake phase is to agree on a certain modulation and coding scheme. By means of the first RTS/CTS exchange the neighboring stations of the sender and the receiver are informed to not use the signalling channel until the whole handshake phase is over. After the second CTS packet is send the neighbors are also informed about the chosen dedicated channel and its reservation time. Two further possibilities were mentioned for channel assignment. The first one is the static assignment, where only one out of N possible channels is used. This approach yields high connectivity due to the related node density. Furthermore
Our investigation focuses on Orthogonal Frequency Division Multiplex (OFDM). OFDM has been chosen as the access technique for the latest WLAN standard IEEE802.11a. Each IEEE802.11a channel has 52 sub–carriers, where four sub– carriers are used for pilot signals while the others convey data. IEEE802.11a uses multiple modulation schemes in combination with different coding rates. For modulation, BPSK, QPSK, 16-QAM and 64-QAM are used. Coding rates are 1/2, 9/16, and 3/4. The combination of coding rates and modulation leads to multiple data rates starting at 6 Mbit/s up to 54 Mbit/s (see also [10]). We assume that eight parallel channels are available referring to the channel spectrum available in the US for IEEE802.11a [11] in the lower and medium band. The maximum distance at which two nodes can communicate (at the lowest achievable data rate of 6 Mbit/s) is around 75 m. A wireless Network Interface Card (NIC) is only able to receive or send on one channel simultaneously. However, we assume that packets can be sent and received over different channels. The NIC needs some time to adjust from one channel
to another. It has to be done in a timely fashion for higher performance. In case two NICs are available, one NIC is for monitoring the common signaling channel only and the second is transmitting data over a dedicated channel. In this case the switching is inherently no problem. In case a packet has to be transmitted over forwarding nodes (thus multi hop is used) some sort of routing has to be applied. Although other more sophisticated protocols could be used, as a preliminary step we use shortest distance routing for all three approaches.
The metric for our performance evaluation are the packet delivery ratio and the transmission delay. Both values are measured at the receiving entity. The delivery ratio is calculated dividing the successfully received packets by the overall amount of packets. The transmission delay is the time needed from the sender to the receiver including the retransmissions. IV. R ESULTS In the following we present our performance evaluation for the given scenarios. We compare our new MAC protocol (labeled dynamic in the figures) with an approach which uses the standard MAC protocol (labeled static in all figures) and an approach that selects the channel randomly (labeled random in all figures). A. Scenario A
Fig. 2.
Frame Trace Friends2x01.
We assume 60 WTs in a surface area of variable size which varies from 30x30 to 500x500 meters. In the Scenario A 10 out of 60 WTs transmit video content to different receiving WTs. Thus, 40 WTs are only forwarding traffic. In the Scenario 1R 10 WTs send to one dedicated node (transition to the wired network) using the other terminals to forward their packets. In Scenario 1S and Scenario 5S each access point transmits to one terminal at a time using the multi–hop capabilities over neighboring terminals if necessary. Note that the traffic load differs for the four scenarios. Each source (WT or access point) selects randomly a trace file from the second season of Friends available at [12] and presented in [13]. 24 different trace files are available. All video sequences have a frame rate of 25 fps, a video format of 400x288 pixels, and a mean bit rate around 750 kbit/s. We refer to [12] for further information. For all video sequences variable bit rate encoding was used as can be seen in Figure 2. Using a frame aggregation level of 800 for better illustration, the frame size versus time is depicted. In [13] we have already shown that the peak rate is much higher than the mean bit rate. For the Friends episodes the peak to mean values are between 7 and 22. This kind of traffic imposes a challenging situation for the multi–hop network. The sending node chooses the video length (no longer than 20 min as this is the maximum length of an episode) and the starting phase randomly between 60 and 100 s. Random starting phases and lengths are chosen to avoid correlation among the traffic sources.
In Figure 3 we give the delivery ratio versus the surface area for the static, random and the dynamic channel assignment. It can be seen that the dynamic approach yields better results than any other approach. The performance of the random approach degrades dramatically as the surface area increases. This is caused by the physical limitation of the network. Only terminals using the same channel can build a multi hop network. In Figure 4 the delay values for the three approaches are given. The best values are achieved by the random approach. But we have to note that the delay values are only calculated for successfully received frames. That means, for the random approach, that if a packet arrives, it will arrive with low delay values. But also the dynamic approach yields good delay values below 100 ms. The static approach has delay values larger than one second if the surface area increases over 70 m. This is due to the fact that for these values the network starts to use multi hop capabilities and the overall network traffic increases. In such a case the limitation to one channel becomes a bottleneck and therefore the delay values increase. The dynamic approach is capable to handle the increase in network traffic using the dynamic assignment of multiple channels. B. Scenario 1R In Figures 5 and 6 the delivery ratio and the mean transmission delay for Scenario 1R realizing video services versus surface area for static, random and dynamic channel assignment is given. The delivery ratio for the dynamic approach outperforms the static one for surface area larger than 150x150 m2 . For smaller surface areas the static approach yields slightly better results due to the absence of the hand shake phase. The random approach results in delivery ratios smaller than 20%. The mean transmission delay given in Figure 4 and Figure 6 are similar. The only change is the delay curve for the dynamic approach increasing for larger surface areas. Still the values are smaller than for the static approach, but for surface areas of 300x300 m2 the delay values are over one second. This makes it hard to realize video services. The increasing delay is caused by a higher number of hops needed to convey a data packet from the access point to its final destination.
Scenario: Ad Hoc
Scenario: Ad Hoc
100
1000
100
delivery delay [s]
frame delivery ratio [%]
80
60
40
20
10
1
0.1
0.01
0
0.001 0
50
100
150
static
200
250 300 size [m*m]
350
random
400
450
500
0
dynamic
50
100
150
static
200
250 300 size [m*m]
350
random
400
450
500
dynamic
Fig. 3. Delivery ratio for video services versus surface area for static, random Fig. 4. Mean transmission delay for video services versus surface area for and dynamic channel assignment. static, random and dynamic channel assignment.
Scenario 1R: One Sink
Scenario 1R: One Sink
100
1000
100
delivery delay [s]
frame delivery ratio [%]
80
60
40
20
10
1
0.1
0.01
0
0.001 0
50
static
100
150
200
250 300 size [m*m]
random
350
400
450
500
dynamic
0
50
static
100
150
200
250 300 size [m*m]
random
350
400
450
500
dynamic
Fig. 5. Delivery ratio for video services versus surface area for static, random Fig. 6. Mean transmission delay for video services versus surface area for and dynamic channel assignment. static, random and dynamic channel assignment.
C. Scenario 1S In Figure 7 and 8 the delivery ratio and the mean transmission delay for Scenario 1S realizing video services versus surface area for static, random and dynamic channel assignment is given. Here as a matter of fact, the produced traffic is only a tenth of the traffic compared to Scenario A and Scenario 1R. Here the static and the dynamic approach yield nearly the same results for both, the delay and the delivery ratio. Only the random approach is suffering due to its channel assignment. D. Scenario 5S In Scenario 5S we increase the traffic by a factor of five. In Figure 9 we see that the static assignment leads to slightly worse values than the dynamic approach. Still the random approach results in very bad delivery ratios. The mean transmission delay given in Figure 10 for the dynamic approach is always lower than 100 ms, while the static approach comes along with delay values of one second if the surface area is larger than 200x200 m2 .
V. C ONCLUSION AND O UTLOOK
In this paper we have once more motivated the interworking between WLAN and 3G networks focusing on low cost WLAN infrastructure. We have introduced a MAC protocol for WLAN multi hop networking, that allows installing WLAN network at low costs. By means of the performance evaluation we have shown that our new proposed MAC scheme can be used for pure ad hoc scenarios as well as scenarios with an infrastructure. Even if the MAC protocol was designed for pure ad hoc communication, it outperforms existing solution even for infrastructure based networks. The performance of such a multi–hop enabled network is satisfactory for demanding multimedia services in terms of delivery ratio and transmission delay. We believe that this will open the door for cost effective deployment of WLAN systems complementing 3G networks and their successors.
Scenario 1S: One Source
Scenario 1S: One Source
100
1000
100
delivery delay [s]
frame delivery ratio [%]
80
60
40
20
10
1
0.1
0.01
0
0.001 0
50
100
150
static
200
250 300 size [m*m]
350
random
400
450
500
0
dynamic
50
100
150
static
200
250 300 size [m*m]
350
random
400
450
500
dynamic
Fig. 7. Delivery ratio for video services versus surface area for static, random Fig. 8. Mean transmission delay for video services versus surface area for and dynamic channel assignment. static, random and dynamic channel assignment.
Scenario 5S: Five Sources
Scenario 5Sg: Five Sources
100
1000
100
delivery delay [s]
frame delivery ratio [%]
80
60
40
20
10
1
0.1
0.01
0
0.001 0
50
static
100
150
200
250 300 size [m*m]
random
350
400
450
500
dynamic
0
50
static
100
150
200
250 300 size [m*m]
random
350
400
450
500
dynamic
Fig. 9. Delivery ratio for video services versus surface area for static, random Fig. 10. Mean transmission delay for video services versus surface area for and dynamic channel assignment. static, random and dynamic channel assignment.
VI. ACKNOWLEDGMENT The work was partially supported by ALCATEL, France. We would like to thank Gerrit Schulte for the fruitful discussions and his help in designing Figure1. Furthermore, we would like to thank all anonymous reviewers for their valuable comments. R EFERENCES [1] Technical Specification Group Services and System Aspects, “3GPP system to Wireless Local Area Network (WLAN) Interworking; System Description (Release 6),” Tech. Rep. V2.4.0, 3rd Generation Partnership Project, Jan 2004. [2] F. H. P. Fitzek, D. Angelini, G. Mazzini, and M. Zorzi, “Design and Performance of an Enhanced IEEE802.11 MAC Protocol for Ad Hoc Networks,” in VTC2003, 2003. [3] F.H.P. Fitzek, D. Angelini, G. Mazzini, and M. Zorzi, “Design and Performance of an Enhanced IEEE802.11 MAC Protocol for Ad Hoc Networks and Coverage Extension for Wireless Networks,” IEEE Wireless Communications, vol. 10, no. 6, December 2003. [4] C. Zhu and M.S. Corson, “A Five-Phase Reservation Protocol (FPRP) for Mobile Ad Hoc Network,” in IEEE INFOCOM 98, 1998. [5] C. Zhu and M.S. Corson, “A Five-Phase Reservation Protocol (FPRP) for Mobile Ad Hoc Network,” Wireless Networks, pp. 371–386, 2001.
[6] S. Mao, S. Lin, S.S. Panwar, and Y. Wang;, “Reliable transmission of video over ad-hoc networks using automatic repeat request and multipath transport,” in 54th IEEE Vehicular Technology Conference, Oct 2001, vol. 2, pp. 615 – 619. [7] K. Ban and H. Gharavi, “Video transmission for multi-hop networks using IEEE 802.11 FHSS,” in International Conference on Image Processing, Sept 2002, vol. 1, pp. 13–20. [8] S. Lin, Y. Wang, S. Mao, and S. Panwar, “Video transport over ad-hoc networks using multiple paths,” in IEEE International Symposium on Circuits and Systems, May 2002, vol. 1, pp. 57–60. [9] H. Li and D. Yu, “Comparison of ad hoc and centralized multihop routing,” in The 5th International Symposium on Wireless Personal Multimedia Communications, Oct 2002, vol. 2, pp. 791–795. [10] Benny Bing, Wireless Local Area Networks, WILEY, 2002. [11] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications – High-speed Physical Layer in the 5 GHz Band, IEEE Standard for Information technology. [12] Reisslein, Fitzek, and Seeling, “Video Traces for Network Performance Evaluation,” http://trace.eas.asu.edu. [13] F.H.P. Fitzek, M. Zorzi, P. Seeling, and M. Reisslein, “Video and Audio Trace Files of Pre-encoded Video Content for Network Performance Measurements,” in IEEE Consumer Communications and Networking ”Consumer Networking: Closing the Digital Divide”. IEEE, January 2004, Las Vegas, Nevada, USA.
