2013 IEEE 24th International Symposium on Personal, Indoor and Mobile Radio Communications: Mobile and Wireless Networks
Coordinated Tethering over Cellular Networks: Handover Scenarios and Signaling Nadine Akkari1, Ghadah Aldabbagh2, Michel Nahas3, Budoor Bawazeer4, John Cioffi5, Haleh Tabrizi6 1,2,4
Faculty of Computing and Information Technology, Computer Science Department, King Abdulaziz University 5,6 Department of Electrical Engineering, Stanford University 1
[email protected],
[email protected],
[email protected],
[email protected],
[email protected],
[email protected].
Abstract—As next generation networks are evolving toward providing the mobile users with higher bandwidth and better Quality of Service, relaying was considered as a solution that will contribute to a better service provisioning to the end users. In a recent paper, grouping the mobile nodes into clusters was proposed as an alternative solution to relaying. This technique allows a coordinated tethering of the wireless nodes in cellular networks controlled by the base station (BS). In this context, for better spectrum usage, some users will relay the information between the BS and other devices using the newly vacated TV white space (TVWS) bands. However, the proposed clustering approach did not take into account the numerous changes that can occur in the cellular network. Therefore, in this paper, we will extend the existing clustering configuration to consider mobile users’ requirements and network events by studying the corresponding handover scenarios and signaling schemes. This will reduce the complexity of the clustering algorithm in terms of time and signaling, and it will improve the network performance and service quality as will be also shown via several user case scenarios. Keywords-Cellular networks, LTE, TV white space, Clustering, Tethering, Handover management
I. INTRODUCTION Relaying has emerged as an effective solution to enhance the performance of new generation of cellular networks. By introducing some relays in a cell, we intend to extend the coverage and improve the throughput and reliability of the network [1]. This will enhance the users’ service provisioning, especially for cell-edge mobile [2]. Initially, relay antennas are distributed within the macro cell but with lower processing capabilities than the Base Station (BS) itself [3]. Most of the research works related to relay channels have considered installation of fixed relay stations in infrastructurebased wireless networks [3] [4]. This is because the deployment of relays in fixed-position antennas is easier than on mobile terminals. However, recent studies have investigated moving relay nodes by considering the users’ mobile stations as mobile relays [5] [6] [7]. A performance evaluation of network architecture with mobile relay stations was realized in [6] and compared with fixed relay stations in [5]. Furthermore, [7] has investigated through field testing, the performance benefits and improvements of multi-hop cellular networks using mobile relays over traditional cellular systems. In addition, papers [8] and [9] have shown potential
978-1-4577-1348-4/13/$31.00 ©2013 IEEE
improvement of handover (HO) performance in cellular systems with the use of relay nodes. Despite their interesting properties, relay nodes require the employment of additional frequency channels. Recently, to combat the spectrum shortage, researchers and industry are searching for new possible frequency bands to be used in cellular networks. Some attractive options are the lately vacated television bands and the unused broadcast television spectra available between 50 and 700 MHz referred to as TV white-space (TVWS) [10] [11]. Furthermore, this frequency range offers better radio propagation properties than frequencies utilized in LTE, because it occurs in lower band. The authors in [10] and [11] have investigated the potential gains of deploying cellular networks in the TVWS band. Furthermore, mechanisms for LTE coexistence in TVWS that dynamically share the spectrum with other secondary users were proposed in [12]. And in [13], the authors studied several use case scenarios of LTE deployment in TVWS spectra in Europe. On the other hand, tethering was lately used over WiFi to grant internet access to a user via another connected device. Tethering became achievable in cellular networks with the abundance subsistence of mobile devices like smartphones, tablets and laptops. Recently, the authors in [14] have proposed a coordinated tethering algorithm to deliver data to a destination node (called slave) through an intermediate nearby device (called hotspot). The proposed algorithm groups iteratively the connected nodes in a cell into clusters. Each cluster consists of a mobile hotspot connected to the BS via LTE. In a cluster, they may exist some slaves attached to the hotspot and communicating with it over TVWS. The hotspot serves as a relay to provide the connected slaves with broadband access to the BS. Because coordinated tethering make use of existing devices and thus does not require any additional infrastructure in the network, it provides a more flexible and affordable solution than relaying. However, in [14], the nodes’ mobility was not taken into consideration. Moreover, the authors have not indicated how to manage the different networks event that can occur in the cell like the arrival of new users or the disconnection of a hotspot. Due to its iterative nature, the clustering algorithm would take a prohibitive time for real-time communications. Therefore, it will not be possible to launch the clustering procedure at each time a network event take place. In this paper, based on the tethering framework proposed in [14], we study the possible network events that will arise in
2170
the tethering context. In the proposed handover scenarios, a mobile node, being a slave or a hotspot, will not undergo the clustering algorithm but will simply perform the handover within the clustering cycles. The handover will provide the mobile node with the “always-on” connectivity with less energy consumption and better Quality of Service (QoS) considerations. The paper is organized as follows. In Section II, the system model and the proposed algorithm for clustering in [14] are presented. The network events that will trigger a handover are described in Section III with the different handover scenarios and signaling schemes. Then, several use case scenarios are given in Section IV to show the merit of the proposed HO procedures. Finally, we conclude in Section V. II. SYSTEM MODEL AND CLUSTERING ALGORITHM We consider a cell of an LTE network with several mobile users inside. After applying the clustering algorithm, the nodes will be divided into hotspots and slaves. The network model is shown in Figure 1. The BS will communicate with the hotspots over LTE allocated frequencies. And each hotspot will be linked with its corresponding slaves using the TVWS bands. In this way, clusters that are comparable to small cells, are formed inside the initial cell as can be seen in Figure 1. Moreover, each node is assumed to be able to operate on both LTE and TVWS bands with the same cellular access technology.
their respective slaves and verifying the feasibility of the configuration in terms of resource allocation. The algorithm consists of five main steps. In the first step, the clustering configuration of the nodes is determined such that it minimizes the overall network transmission power. The clustering is performed by the BS based on the average channel gains of different BS-nodes and nodes-nodes links. Then, the remaining four steps aim at verifying that the clustering is feasible with the available resources while preserving the service requirements of different users like power and rate. If the configuration prove to be not possible, then step 1 (clustering) is performed again taking into consideration the nodes that did not meet their requirements. Further details can be found in [14]. However, the clustering algorithm presented above has some drawbacks. First, it is somehow complex and required a large amount of time to be completed. This is because several iterations may be needed to determine the different clusters. Also, at each iteration, many optimization problems are performed to verify the different constraints on resource allocation and on power and rate requirements in the clusters. Moreover, this proposed method does not take into consideration important cellular events like the mobility of the nodes, the arrival of a new node or a node leaving the cell. To take care of such changes, one solution is to perform a reclustering periodically and more often. But, this solution is costly because of the long time it would take. Therefore, here, we propose to detect any network change and to solve it without the need to trigger a new clustering. This way, the clustering procedure can be performed more rarely which will save a considerable time and signaling in the network. Furthermore, by responding more quickly to the requirement of the users in terms of bandwidth and power, the overall system will have a better QoS. III. NETWORK EVENTS AND HANDOVER SIGNALING In this section, the different handover cases between the clusters will be explained. The possible HO scenarios and the network events are examined with their respective signaling.
Fig. 1. Clusters formed in a macro cell.
Grouping the nodes into clusters will decrease the total transmission power and allow frequency reuse among the clusters. Moreover, the cell coverage is enlarged by reducing the path-loss while choosing the hotspots closer to the slaves than the BS. Also, the network capacity can be expanded: a single BS in the cell can serve more users with the help of the hotspots connected to their slaves over TVWS frequencies. The clustering algorithm proposed in [14] will be explained briefly hereafter. Actually, the algorithm will iteratively cluster the nodes by identifying the hotspots with
A. Handover Scenarios: After the clustering process, a mobile node could move from one cluster to another by performing a horizontal handover. In this case the slave will move from one TVWS cluster to another TVWS cluster. This is due to a node detecting a low Received Signal Strength (RSS) and will hence decide to move to another cluster. Also, the mobile node could be requesting higher bandwidth compared to the current available one and as a result moving to another cluster could happen or a direct LTE connectivity could be a better choice. In this case, the mobile will change its role from a slave to a hotspot. Thus, the handover triggering parameters could be either the RSS or the bandwidth requirements. We could depict three different scenarios: - Scenario 1: Slave becomes a hotspot (TVWS to LTE). - Scenario 2: Slave of hotspot becomes a slave to another hotspot (TVWS to TVWS). - Scenario 3: Hotspot becomes a slave of another hotspot (LTE to TVWS).
2171
Depending on the type of network interface involved in the handover process, we can classify previous scenarios into horizontal handover (same interfaces) or vertical handover (different interfaces). Since the node in Scenario 1 will move from TVWS to LTE and in Scenario 3 will move from LTE to TVWS, they are considered as vertical handover (VHO). Scenario 2 will be considered as horizontal handover (HHO). Different network events can initiate a handover. Table I lists these different activities occurring in a cellular network after clustering. The table shows also the possible handovers that take place and the nodes that will be affected in this change. For instance, when a node moves inside the cell (Events 1 and 2 in Table I), its RSS may change and become below a certain threshold. In this case, the node should change its cluster. Therefore, a slave will be related to another hotspot (HHO) or it may also become a hotspot (VHO). It can be noticed from the table that, depending on the event, one or more scenario may apply and that these scenarios cover all the possible events. In the case of a new arriving node to the cell (Event 5), it will be connected to the BS directly as a hotspot without any slave initially but other nodes can join its cluster after a HO. Next, the signaling for the three scenarios mentioned above is considered in details. Table I. Network events with corresponding HO and nodes Network Event
1. Slave moves
2. Hotspot moves
3. Slave needs more bandwidth
4. Hotspot needs more bandwidth
Possible HO - HHO (TVWS to TVWS) - VHO (TVWS to LTE) - HHO (TVWS to TVWS) - VHO (LTE to TVWS) - HHO (TVWS to TVWS) - VHO (TVWS to LTE) - VHO (TVWS to LTE) - HHO (TVWS to TVWS)
Involved Nodes BS, the slave, the old hotspot and the new hotspot BS, the hotspot and all the slaves within its cluster BS, the slave, the old hotspot and the new hotspot BS, the hotspot and all the slaves within its cluster
5. New node becomes active
- New LTE band
BS and the new node
6. Slave becomes inactive
- No HO needed
BS and the hotspot of the slave
7. Hotspot becomes inactive
- VHO (TVWS to LTE) - HHO (TVWS to TVWS)
BS, the hotspot and all the slaves within its cluster
B. Scenario 1: VHO (TVWS to LTE) The proposed handover signaling for VHO, shown in Figure 2, is based on [15] where the main signaling messages in the context of LTE handover are presented. In LTE HO, the mobile node should process the handover measurements and send the results to the BS to make a handover decision.
Fig. 2. VHO: slave changes to be a hotspot.
In our case, the messages exchanged are within the same LTE cell and under the same BS. Thus, the HO signaling process will be performed as follows: • At first, data are exchanged between BS and slave through the hotspot node as per (1). The hotspot is acting as a relay node. • The Measurement report (2) is considered as the base of all steps that come later. Depending on the Measurement report that includes values for signal strength, bandwidth requirement, and other network parameters, the hotspot will decide whether or not the node will perform the handover. The Report is sent from the slave node to the hotspot to take the handover decision. • The “HO Decision” (3) will depend on the report values without specifying the target of the handover process. At this stage a “Handover Request” message (4) is sent from the hotspot to the base station. • The BS is responsible to find the target network to perform the handover to. There are two possible options: first, the selected target network could be another cluster, thus the slave will be connected to another hotspot by performing Horizontal Handover (HHO). Second, the slave could be connected directly to the BS (VHO). In this case, the VHO is performed from TVWS to LTE as per our current scenario. • At the “Admission control” phase (5), the target will make decision whether to accept the request or to reject it. In this current scenario, BS will perform the admission control. If the request is accepted, BS stops sending any new data to the hotspot and start buffering until handover is performed. • Next, the “Handover Request Ack” (6) is sent from BS to the hotspot as a response to the handover request message. The message contains all the information about the new link between BS and slave node and the link on which to forward all non-received data by the slave node. • “RRC Connection Reconfigure message” (7) will be sent from BS to the slave node directly right after the
2172
• •
• •
“Handover Request Ack” message, since BS controls all the nodes information inside the cell. This message contains all the required information for the slave node to perform the VHO handover. At stage (8), the hotspot will forward buffered data to BS which will buffer incoming packets (9) until the new connection is established. Synchronization (10) is performed between the slave node and the BS for the slave to access the BS before the “RRC Connection Reconfiguration Complete” message (11) is sent from the slave node to the BS to confirm the handover. The BS in turn will update the lists of hotspots and related slaves. Packet-data will be sent on the new link (12) as the BS starts to send data to the slave which is now a hotspot. “UE Context Release” is sent form the BS to the old hotspot (13) to release all the recourses which were allocated to the old slave. At the hotspot node, all resources related to the old slave will be freed now by performing “Release resources” (14).
C. Scenario 2: HHO (TVWS to TVWS) If the BS decides that it is better to change the slave's cluster rather than to connect to it directly as a hotspot, horizontal handover will occur. Figure 3 shows HHO signaling.
Then, the source hotspot forwards the data to the BS. And the BS will send them to the target hotspot that will buffer packets until connection is established. The slave now perform synchronization to access the target hotspot and send “RRC Connection Reconfiguration Complete” to the target hotspot after accessing it to confirm the handover. The lists of hotspots and related slaves should be updated. Finally, the target hotspot will start sending data to the slave. The “UE Context Release” message will be sent from new hotspot to the old hotspot via the BS and all related resources will be freed. D. Scenario 3: VHO (LTE to TVWS) When the node which needs to perform the handover to TVWS is a hotspot, the same messages as per the previous scenarios will be exchanged between the target hotspot and BS. HO Decision will occur at the BS and the handover request message will be sent to the target hotspot. Similarly, the target hotspot will accept or reject the request. IV. USER CASE SCENARIOS Consider a dense mobile area modeled by a 100 × 100 meter square field crowded with 24 active users. The BS, located at the center, operates based on the LTE standard with 10 resource channels of 360 KHz each. After applying the clustering algorithm, the users will be divided into clusters and each node is assigned to be a hotspot or a slave (Figure 4). The users are numbered from 1 to 24 and the hotspots are indicated by the letter “h” in a red rectangle. The remaining nodes are slaves and they are connected to their respective hotspot over available broadcast television band (TVWS). For instance, in Figure 4, node 1 is a hotspot and it has four slaves: nodes 10, 13, 19 and 24. Node 20 is a hotspot with only one slave node 18. Node 15 does not have any slave. Clustering
50
23
meters
9 0
16
11 7
17 h4
h15
h21
h20
BS
12
h1 h14
5
8 6
At first, the same flow of messages will occur as per the case of VHO before. For the “Handover Request” message, it will be sent from hotspot to the BS which will find the better target to perform the HO to. In this scenario, BS decides to change the slave’s cluster, so HHO from TVWS to TVWS will be performed. Thus the BS will send the handover request to the selected hotspot. In this case, the target hotspot will make the decision whether to accept the request or to reject it. A “Handover Request Ack” message will be sent from the target hotspot to the BS. Then, the BS will send the HO Request Ack to the source hotspot. As per Scenario 1, the “RRC Connection Reconfigure” is sent directly from BS to the slave node since BS controls all nodes inside the cell.
-50 -50
18
19
3
Fig. 3. HHO: slave changes its cluster.
22
h2
13
24 10
0 meters
50
Fig. 4. Clustering model for user case scenarios.
A. Node moving: Assume that node 19 is moving away from node 1 and approaching from node 20. In this case, the Received Signal Strength (RSS) from node 1 will deteriorate and drop below a certain threshold. The communication quality will become unacceptable. Here, a HO request will be sent to node 20 and node 19 will change its cluster and become a slave of node 20 (HHO: TVWS to TVWS).
2173
B. Bandwidth needed: Now suppose that node 3 is requesting more bandwidth because it is running a highly demanding application like video streaming for instance. Then, if the hotspot (node 21) cannot afford this increasing bandwidth of node 3, a HO request will be sent to node 14. Suppose that also node 14 is not able to answer the need of node 3 in terms of bandwidth, then node 3 will become a hotspot and will be directly connected to the BS via LTE band (VHO: TVWS to LTE). From the use case scenarios, we can deduce that the slave node will either perform a HHO or a VHO based on the user's needs in terms of RSS and bandwidth. In both cases, clustering will not be triggered. However the handover will be performed as per the current network conditions and user's requests. C. Number of handovers: Next, the number of VHO and HHO occurring in the cell is calculated in function of the average velocity of the nodes. For this, we vary the average velocity of the nodes from 1 to 10 meters per second and then compute for each velocity the average number per minute of VHO and HHO. The maximum allowed distance between any 2 connected devices (BShotspot or hotspot-slave) is 50 meters. From Figure 5, it can be seen that the number of Handovers increases with the average velocity as expected. Moreover, more HHO take place than VHO meaning that the slaves would change their cluster more than switching to become hotspots. Average number of Handovers vs average node velocity
35
Average HO number per minute
25
REFERENCES [1] R. U. Nabar, H. Bolcskei, and F. W. Kneubuhler, "Fading relay channels: performance limits and space-time signal design," IEEE Journal on Selected Areas in Communications, vol. 22, pp. 1099-1109, Aug. 2004. [2] J. N. Laneman, D. Tse, and G. W. Wornell, "Cooperative Diversity in Wireless Networks: Efficient Protocols and Outage Behaviour," IEEE Transactions on Information Theory, vol. 50, no. 12, Dec. 2004. [3] Y. Liu, R. Hoshyar, X. Yang, and R. Tafazolli, "Integrated radio resource allocation for multihop cellular networks with fixed relay stations," IEEE Journal on Selected Areas in Communications, vol. 24, pp. 2137-2146, Nov. 2006. [4] A. Adinoyi and H. Yanikomeroglu, "Cooperative relaying in multiantenna fixed relay networks," IEEE Transactions on Wireless Communications, vol. 6, no. 2, pp. 533-544, Feb. 2007. [5] H. Nourizadeh, S. Nourizadeh, and R. Tafazolli, "Performance Evaluation of Cellular Networks with Mobile and Fixed Relay Station," in Proceedings of IEEE Vehicular Technology Conference (VTC Fall), Montreal, Canada, 25-28 Sept. 2006. [6] M.F.M. Hossain, A. Mammela, and H. Chowdhury, "Impact of Mobile Relays on Throughput and Delays in Multihop Cellular Network," in Proccedings of International Conference on Wireless and Mobile Communications (ICWMC), Athens, Greece, July 27 - Aug. 1, 2008. [7] J. Muñoz, B. Coll-Perales, and J. Gozalvez, "Research testbed for field testing of Multi-hop Cellular Networks using Mobile Relays," in Proceedings of IEEE Conference on Local Computer Networks (LCN), Denver, United States, 10-14 Oct. 2010.
[9] J. Kim and D.-H. Cho, "Pre-buffering Scheme for Seamless Relay Handover in Relay Based Cellular Systems," in Proceedings of IEEE Vehicular Technology Conference (VTC-Spring), Barcelona, Spain, 2629 April 2009.
20
15
[10] T. Dudda and T. Irnich, "Capacity of cellular networks deployed in TV White Space," in Proceedings of IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), Washington, United States, 16-19 Oct. 2012.
10
[11] H. Fujii, S. Miura, and H. Kayama, "Novel cognitive radio technique for using white space in public cellular networks," in Proceedings of 18th Asia-Pacific Conference on Communications (APCC), Jeju Island, Korea, October 15-17, 2012.
5
0
This paper was funded by the Deanship of Scientific Research (DSR), King Abdulaziz University, under grant No. (11-151432 HiCi). The authors, therefore, acknowledge with thanks DSR technical and financial support.
[8] B. Coll-Perales and J. Gozalvez, "On the Capability of Multi-hop Cellular Networks with Mobile Relays to Improve Handover Performance," in Proceedings of International Symposium on Wireless Communication Systems, Aachen, Germany, 6-9 Nov. 2011.
Number of VHO Number of HHO Total number of Handovers
30
ACKNOWLEDGEMENT
1
2
3
4
5
6
7
8
9
10
Velocity [m/s]
[12] M. Beluri et al., "Mechanisms for LTE Coexistence in TV White Space," in Proceedings of IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), Washington, United States, 16-19 Oct. 2012.
Fig. 5. Average HO number for different nodes’ velocity.
V. CONCLUSION This paper extended the newly proposed clustering approach that assigns some users to act as hotspots by relaying the BS information to other users (slaves) based on a relatively complex algorithm. As the clustering procedure did not take into account the changes occurring in a mobile environment, we have identified the different network events that could generate different handover scenarios. We have studied the related signaling schemes and presented some user case scenarios which will improve the service quality compared to the clustering algorithm with no such considerations. Furthermore, a detailed performance evaluation is being considered as an extension to this work.
[13] V. De Beek, J. Riihijarvi, and P. Mahonen, "Intrinsic challenges (and opportunities) to deploy LTE in Europe’s TV white spaces," in Proceedings of Future Network and MobileSummit Conference, Berlin, Germany, 4-6 July 2012. [14] H. Tabrizi, G. Farhadi, and J. Cioffi, "Tethering Over TV White-space: Dynamic Hotspot Selection and Resource Allocation," in Proceedings of IEEE Vehicular Technology Conference (VTC Fall), Las Vegas, United States, 2-5 Sept. 2013. [15] N. Nasser, A. Hasswa, and H. Hassanein, "Handoffs in fourth generation heterogeneous networks," Communications Magazine, IEEE, vol. 44, no. 10, pp. 96-103, Oct. 2006.
2174
