A Combined Vertical Handover Decision Metric for QoS Enhancement in Next Generation Networks Anna Maria Vegni 1, Gabriele Tamea 2,Tiziano Inzerilli 2 and Roberto Cusani 2 Abstract Vertical handover (VHO) techniques are applied when mobile users require service continuity and a seamless roaming between heterogeneous networks. A VHO decision can be taken on the basis of physical parameters, such as the received signal strength (RSS), data rate and signal-to-interference and noise ratio (SINR). In this paper, we compare three VHO decision criterions, each of them based on different physical metrics. A smart combination of several VHO criterions is the core of our proposed approach, which uses not only RSS or SINR parameters, but also a hybrid mixture from different wireless access networks. To improve quality of service for mobile users, a data rate gain parameter is introduced that lets an efficient VHO be executed. Simulation results of the proposed VHO approach are in terms of both end-users efficiency, i.e. cumulative received bits, and network performances, i.e. VHO frequency. Keywords- vertical handover; mobility management; data rate gain; VHO frequency. I. INTRODUCTION On next generation wireless networks heterogeneous broadband technologies coexist in order to guarantee a seamless connectivity to mobile users. Different network characteristics are basically expected to different multimedia applications [1]. So, as mobile applications require service and quality-of-service (QoS) continuity, cooperation of mobileaccess networks in heterogeneous environments is an important feature to assure [2]. In this context, vertical handover (VHO) techniques can be applied when connectivity switching is needed to preserve host connectivity and optimize QoS as perceived by the end user. A VHO is a process preserving user s connection on the move and following changes of network, (i.e. from a base station (BS) UMTS to an access point (AP) WIFI) [3]. Many approaches are used to vertical hand over from a serving network (SN) to a candidate network (CN), with the goal of provide a maximization of throughput, and a limitation of unwanted and unnecessary vertical handovers [4]. This aspect is also called ping-pong effect [5] and leads to excessive network resource consumption and also affects mobile terminal s performance (i.e. battery life). As a consequence, the minimization of the number of VHOs is an important deal in handover management and many criterions 1 A.M. Vegni is with the Department of Applied Electronics, University of Roma Tre, Via della Vasca Navale 84, 00146 Rome, Italy. Email: avegni@uniroma3.it. 2 G. Tamea, T. Inzerilli, and R. Cusani are with the Department of INFOCOM, University of Rome Sapienza, Via Eudossiana 84, 00184 Rome, Italy. E-mail: {tamea, inzerilli, cusani}@ uniroma1.it. have been proposed in the state of the art. In [4] the authors present an algorithm based on predicted distance information from a mobile terminal (MT) to a wireless cell. Limitation of the ping-pong effect is just obtained by MT s distance, but no physical network parameters are mentioned as VHO decision criterion. Normally, a handover initiation is typically driven on traditional RSS parameter, though it does not give good performance for MT wide QoS requirements [6]. Also the SINR parameter is particularly used as a metric in order to provide seamless handover with adaptive data rate (DR), as described in [7], while information about MT s location [8] or monetary cost [9] can be used to preventive VHOs. Combination of the above metrics can generate most effective VHO decisions, called as hybrid VHO approaches, and based on two or more metrics, such as the RSS and the distance information between a BS and a MT [10]. In [11] a VHO decision function is obtained as a composition of the monetary cost, power requirements, security parameters, MT preference, network conditions, and MT s speed. No simulation results are reported for this approach. Then, in [7] the authors address on the SINR parameter as VHO metric, but do not consider it as hybrid vertical handover criterion. In this paper, the main physical parameters (i.e. RSS, SINR and DR) are opportunistically mixed in a hybrid VHO approach, in order to improve end-user QoS. A comparison between three different VHO decision functions (VHDFs), based on single and combined VHO decision metrics, has been evaluated. Single metrics are respectively based on RSS and DR, while the combined one is a contribution of RSS plus SINR and DR parameters. In Section II we introduce both the single-metric based VHDFs (i.e. called as RSS-VHDF, and DR-VHDF), while in Section III the combined-metric based VHDF (C-VHDF) is proposed. Effectiveness of C-VHDF is tested in terms of maximization of cumulative received bits and limitation of number of VHOs (Section IV). Finally, the main findings of our study are summarized in Section V. II. SINGLE-METRIC BASED VHDF We present two VHDFs based on single metrics, such as the RSS and DR parameters, respectively illustrated in Subsection A, and B. A. RSS-VHDF The RSS-VHDF is based on RSS parameter, measured by different MT s Network Interface Cards (NICs), (i.e. UMTS, and WIFI NICs) [6]. Basically, smartphones and mobile devices are equipped with several NICs, such as UMTS, GSM,
WIFI, Bluetooth, GPS, and so on. Let us assume that a MT is moving in an outdoor heterogeneous network environment, composed by different network hotspots (i.e. UMTS macrocells and WIFI microcells). The MT moves at low speed (i.e. 0.5 m/s, depicting a pedestrian environment) and in each position it can receive the RSS parameter by the SN, and measures the RSS parameter by a CN, if available. Without loss of generality, we can assume that UMTS is the SN, and WIFI a possible CN. The RSS-VHDF performs a power monitoring of the SN at regular intervals of times, called as waiting times, in order to limit both the battery life of the MT and the number of vertical handovers [6, 8]. The main steps of RSS-VHDF are the following: 1. First VHO alarm: when the RSS level from the SN is decreasing under a fixed value (i.e. P SN < S SN + Th 1 ), a first VHO alarm is given by the MT. Basically, S SN is the MT s sensibility for a particular SN, and Th 1 a threshold to avoid unexpected connection interruptions; 2. Network discovery: after the first VHO alarm has occurred, the network discovery is initiated in order to find an available CN to handover to; 3. Power Testing: if a CN is available and its RSS parameter is higher than a VHO threshold (i.e. P CN > H CN-VHO ), the channel estimation phase starts; 4. Channel estimation: the channel estimation is evaluated when the MT requires a QoS enhancement. The goodput parameter is instantaneously estimated (i.e. exponential smoothing average), both for the SN and the CN (i.e GP SN, and GP CN, respectively). If GP CN > GP SN, a VHO is performed; otherwise the MT will repeat this phase later, after waiting T wait seconds [6, 8]; 5. Idle mode: just after a VHO occurrence, the MT is in idle mode for at least the waiting time parameter T wait. The values of T wait are independent on network technology. The RSS-VHDF approach is described in Figure 1. The power testing phase is the core of RSS-VHDF, as a VHO occurs only if the CN s RSS parameter exceeds a VHO threshold, i.e. P CN > H CN-VHO. It represents a necessary condition for VHO execution, and after this a channel estimation is performed [6]. If (1) is not verified, no VHO will be executed. Basically, the RSS-VHDF performs the goodput estimation in order to guarantee a QoS improvement by switching in a CN. The waiting time parameter is introduced to limit the pingpong effect, (i.e. when a MT moves on the boundary of a wireless cell and makes frequent handovers). The waiting time parameter is defined as the minimum time that must be assured between two consecutive handovers [6, 8]. The greater is the waiting time the smaller will be the number of performed vertical handovers, and for all the waiting time duration the MT is in idle mode. We considered the same values for T wait, both if the MT is in the SN, and in the CN, as: T wait,k = 30 + t, (1) Figure 1. Decision tree for RSS-VHDF approach. where t =15k, with k =0, 1,, K, (i.e. K=7). As an example, for k=0, T wait,0 =30, the MT will wait for 30 seconds connected to the SN, and the RSS-VHDF will not work [6, 8]. In RSS- VHDF, ping-pong effect is strongly limited by the VHO waiting times, and the channel estimation. B. DR-VHDF The DR-VHDF approach is based on DR parameter, measured in the SN and the available CN. A VHO occurs when the connectivity switching can assure a data rate gain (i.e. R gain ). This parameter is similar to the goodput estimation in RSS-VHDF, but it represents an instantaneous measurement, and not an estimation [6, 8]. Basically, as the SINR parameter is strictly dependent on the data rate R [Mbps], R gain is obtained by a comparison between data rate parameter in the SN and that in the CN. The SINR parameter in WIFI network represents the ratio between the signal received by the i-th MT from j-th WIFI AP, and that of all other interfering signals plus background noise power at MT receiver end, (i.e. P B ), i.e. as defined in [7]: SINR WIFI j,i = G WIFI WIFI j,i P j, (2) P B + G WIFI WIFI k,i G k kwifi k j where P WIFI j is the transmitting power of j-th WIFI AP, and G WIFI j,i is the channel gain between i-th MT and j-th WIFI AP. Analogously, in UMTS network the SINR parameter is expressed as, [7]: SINR UMTS j,i = P B + kumts G UMTS UMTS j,i P j,i G UMTS UMTS ( k,i P k ) G i, j UMTS UMTS P j,i,(3) where P UMTS k is the total transmitting power of the k-th UMTS BS, P UMTS j,i is the transmitting power of the j-th UMTS BS to
Figure 2. Decision tree for DR-VHDF approach. the i-th MT, and G UMTS j,i is the channel gain between the j-th UMTS BS and the i-th MT. From (2) and (3), it derives the maximum achievable data rate R WIFI,UMTS j,i for the i-th MT connected with the j-th UMTS BS, and WIFI AP, respectively. Then, by considering the downlink transmission from the AP/BS to the MT, respectively for WIFI and UMTS networks, for a given carrier bandwidth W, and by Shannon capacity formula, the data rate expression can be written as, [12, 13] WIFI,UMTS = W log 2 1 + SINR j,i R j,i WIFI,UMTS, (4) where represents the channel coding loss factor [db]. For WIFI and UMTS networks, we consider W WIFI = 22 MHz [12] and W UMTS = 5 MHz [13], while the parameters WIFI and UMTS correspond to 3 db [12] and 12 db [13], respectively. In order to assure service continuity, the data rates from two different networks have to be the same or differ of a bandwidth gain value R gain [Mbps]. As an example, if we consider a handover from UMTS to WIFI, data rate comparison is given by R WIFI j,i R UMTS j,i + R WIFI gain, (5) VHO decision that is based on a data rate gain. Just after the first VHO alarm and the network discovery, like in RSS- VHDF, the DR-VHDF introduces the Data Rate gain phase, as a VHO decision criterion. No power testing phase is considered because the metric is different. The core step of DR-VHDF is the Data Rate gain, whose condition is expressed by (5) and (6), when UMTS is the SN, and WIFI the CN, and vice versa. Briefly, the following steps compose the DR-VHDF algorithm: 1. First VHO alarm: as in RSS-VHDF; 2. Network discovery: as in RSS-VHDF; 3. Data rate gain: if a CN is available, a data rate comparison is performed between SN and CN. If condition (5) or (6) occurs (i.e. depending if WIFI is the CN and UMTS the SN, and vice versa), a VHO is initiated to the CN; 4. Idle mode: as in RSS-VHDF. III. COMBINED-METRIC BASED VHDF In the C-VHDF, the combined metric is based on the power control and the data rate gain; the first is typical from RSS- VHDF, the latter from the DR-VHDF. In this way, a VHO is performed if the power testing is verified according, and then a data rate gain is assured by the CN, as expressed in (5) or (6). The C-VHDF process is depicted in Figure 3. The main steps are: 1. First VHO alarm: common to RSS and DR-VHDF; 2. Network discovery: common to RSS and DR-VHDF; 3. Power testing: common to RSS-VHDF; 4. Data Rate gain: common to DR-VHDF; 5. Idle mode: common to RSS and DR-VHDF. The combined RSS and DR metrics avoid reducing the VHO frequency thought maximize network performance (i.e. cumulative received bits). The comparison of the three VHDF approaches is illustrated in the following Section IV. where R WIFI gain, represents the data rate gain when moving from the UMTS to the WIFI network. Again, in case of a handover from WIFI to UMTS, the data rate condition will be: R UMTS j,i R WIFI j,i + R UMTS gain, (6) where R UMTS gain represents the data rate gain when moving from the WIFI to the UMTS network. Hence, in both types of transitions, a vertical handover occurs only if the network switching can offer a gain of data rate. In Section IV, the data rate gain for VHO from UMTS to WIFI has been set equal to that from WIFI to UMTS, (i.e. R UMTS gain = R WIFI gain ), for different values, in order to be VHO type independent. The overall DR-VHDF process is illustrated in Figure 2. This approach has main analogies with RSS-VHDF, but the Figure 3. Decision tree for C-VHDF approach.
IV. SIMULATION RESULTS A simulation environment is set up in order to validate the effectiveness of the C-VHDF algorithm, with respect to RSS and DR-VHDF. Results were averaged over several heterogeneous scenarios (i.e. 100 scenarios), each of them based on 3 UMTS cells and 30 WIFI APs, modeled in a region of 2 Km 2 by a map of 400400 zones of 5 m 2 [6, 8]. The MT moves for 2500 steps in a pedestrian environment, (i.e. 0.5 m/s speed). In order to approximate a real pedestrian and outdoor environment, each scenario is generated in a random way, such as the AP and BS positions are unknown a priori, and the MT moves in this area in random directions. The probability to perform a handover from UMTS to WIFI and vice versa is strictly dependent by the network setup. Unwanted and unnecessary VHOs can occur as the MT moves in an area with radio coverage (i.e. UMTS and WIFI) unknown a priori. Table I collects the main parameters for WIFI and UMTS cells [6, 12]. According to (2) and (3), the interference power is 40% in WIFI channel, while the ratio of total allocated BS transmits power to UMTS channel is 80%, respectively. Then, the transmitted power in the middle of UMTS and WIFI cell is about 43 dbm and 30 dbm, respectively [14, 15]. For the VHO thresholds H CN-VHO for WIFI and UMTS, we respectively set 94 dbm and 97 dbm, values greater than the sensibility parameter. Finally, the RSS-VHDF performances are obtained by an exponential smoothing average channel bandwidth estimation [6], while for DR and C-VHDF we considered R WIFI/UMTS gain = 0.3, 0.5, 0.7, 0.9, and 1 Mbps. Figures 4, 5 and 6 show the average of CRBs [Bit] at the end of the MT s walk for all three VHDFs, versus different values of waiting time parameters (1). As explained in Section II-A, the RSS-VHDF performs a goodput estimation in order to make a vertical handover. The estimated goodput gain is GP gain and represents the ratio GP CN /GP SN, (i.e. in the simulations we considered 0.3, 0.5, 0.7, 0.9, and 1). The maximization of system performance (i.e. CRBs) shows the effectiveness of C- VHDF algorithm. As a matter of fact, the average of CRBs sampled at the 2500-th step for C-VHDF has low different values, for each value of waiting times and data rate gains (Figure 4). DR and RSS-VHDF have high different values, decreasing for increasing values of waiting time parameter, as shown in Figure 5 and 6, respectively. The C-VHDF gives maximum performances, independently of the waiting time parameter. As the tradeoff between maximization of CRBs and restriction of number of VHOs is an issue for VHO management, the single-metric based VHDFs have lower performance than combined-metric based VHDF. This is shown in Figures 7, 8 and 9 by the average of the number of vertical handovers occurred for the DR, RSS, and C-VHDF vs. the waiting time parameter, respectively. The VHO frequency represents the total amount of VHOs, executed during MT s path. It can be expressed as: TABLE I. Max. Tx Power Cell Radius MAIN SIMULATION PARAMETERS Channel Gain Sensibility Background noise UMTS 43 dbm 600 m 20% 100 dbm 100 dbm WIFI 20 dbm 120 m 60% 100 dbm 96 dbm Figure 4. C-VHDF Cumulative Received Bits. Figure 5. DR-VHDF Cumulative Received Bits. N xvhdf, j = n VHO i ( T wait,k ), (7) i=1 where x-vhdf,j is VHO frequency for RSS, DR, and C-VHDF, for the j-th waiting time T wait,k (i.e. j=k+1, with jj, J=8). Figure 6. RSS-VHDF Cumulative Received Bits.
Then, N is the total number of steps of the MT s path (i.e. N=2500). The value of n VHO i is 1 if a VHO has occurred in the i-th step, 0 otherwise. The higher the VHO frequency the lower the waiting time parameter (i.e. j=1), and vice versa (i.e. j=8). In Table II the maximum and minimum values of x-vhdf,j are collected for RSS, DR, and C-VHDF, respectively. The value of C-VHDF,1 is lower than all the other maximum VHO frequency values for RSS and DR-VHDF. It shows that the C- VHDF strongly reduces the ping-pong effect, but maintains high CRB values. Let us consider the following parameter as an indicator of VHO frequency: x -VHDF = x -VHDF,1 x -VHDF,8. (8) It represents the VHO frequency band, that is the range of all possible values of VHO frequency for a particular VHDF. It is obtained as the difference between the maximum and minimum value of averaged VHO frequency. From Table II, we get that for RSS and C-VHDF the frequency bands are equal to five, i.e. RSS-VHDF = C-VHDF =5, while the DR- VHDF = 32, a very wide band. For the same value of data rate gain (i.e. 0.5 Mbps), C- VHDF has lower number of VHOs than that for DR-VHDF. The C-VHDF algorithm never joins the minimum number of VHOs reached by DR-VHDF, independently of the waiting time parameter. As a matter, the range of VHO frequencies for C-VHDF is around [1, 5], while for DR-VHDF is [13, 45]. This proves the effectiveness of a combined VHO approach, respect to a single criterion-based algorithm, while keeping high network performance. The worst case for C-VHDF is obtained for R gain =0.5 Mbps, as the VHO frequency is around 5, and the CRBs are higher than 8 Gbits. The same value of CRBs is obtained by DR- VHDF for R gain =0.5 Mbps and DR-VHDF,1 = 44; while the RSS- VHDF reaches 8 Gbits for R gain =0.5 Mbps and RSS-VHDF,6 = 8. Figure 7. Average VHO frequency for C-VHDF. Figure 8. Average VHO frequency for DR-VHDF. TABLE II. MAXIMUM AND MINIMUM VALUES OF VHO FREQUENCY RSS-VHDF, j DR-VHDF, j C-VHDF, j j = 1 12 45 5 j = 6 7 13 1 CONCLUSIONS We have presented a comparison between three different VHO decision approaches, i.e. the RSS, the DR, and the C- VHDFs, based on single and combined metrics, respectively. Simulation results show that C-VHDF can assure better network performance in terms of limitation of VHO frequency, respect to single metric-based VHDF algorithm. The main aspect of the proposed C-VHDF algorithm is that a VHO is performed if an available CN can assure a data rate gain. The C-VHDF provides the same performances of RSS and DR- VHDF, in terms of maximization of CRB, but with a reduction of the ping-pong effect. Figure 9. Average VHO frequency for RSS-VHDF.
REFERENCES [1] K. Knightson, N. Morita, and T. Towle, NGN architecture: generic principles, functional architecture, and implementation, IEEE Communications Magazine, Vol. 43, No. 10, pp. 49-56, Oct. 2005. [2] I.F. Akyildiz, et al., Mobility management in next-generation wireless systems, in Proc. of the IEEE, Vol. 87, No. 8, pp. 1374 1384, August 1999. [3] M. Stemm, and R.H. Katz, Vertical Handoffs in Wireless Overlay Networks, ACM/Kluwer Mobile Networks and Applications (MONET), Vol. 3 issue 4, 1998. [4] X. Yan, Y.A. Sekercioglu, and N. Mani, A method for minimizing unnecessary handovers in heterogeneous wireless networks, in Proc. of Int. Symposium on a World of Wireless, Mobile and Multimedia Networks, WoWMoM 2008, pp. 1-5, 23-26 June 2008. [5] W.I. Kim, et al., Ping-Pong Avoidance Algorithm for Vertical Handover in Wireless Overlay Networks, in Proc. of IEEE 66 th Vehicular Technology Conference, VTC-2007 Fall, pp. 1509-1512, Sept. 30 2007-Oct. 3 2007. [6] T. Inzerilli, and A.M. Vegni, A reactive vertical handover approach for WiFi-UMTS dual-mode terminals, in Proc. of 12th Annual IEEE International Symposium on Consumer Electronics (ISCE 2008), CDROM no. of pages: 4, 14-16 Apr. 2008, Vilamoura (Portugal). [7] K. Yang, I. Gondal, B. Qiu, and L. S. Dooley, Combined SINR based vertical handoff algorithm for next generation heterogeneous wireless networks, in Proc. of IEEE Global Telecommunications Conference (GLOBECOM 2007), 26-30 Nov. 2007, Washinton DC, USA. [8] T. Inzerilli, A.M. Vegni, A. Neri, and R. Cusani, A Location-based vertical handover algorithm for limitation of the ping-pong effect, in Proc. of IEEE Int. Conference on Wireless and Mobile Computing, Networking and Communications (WiMob 2008), 12-14 Oct. 2008, Avignon (France). [9] J. McNair, and Z. Fang, Vertical handoffs in fourth-generation multinetwork environments, IEEE Wireless Communications, Vol. 11, Issue 3, pp. 8-15, June 2004. [10] H. Bing, C. He, and L. Jiang, Performance analysis of vertical handover in a UMTS-WLAN integrated network, in Proc. of IEEE Int. Symposium on Personal, Indoor and Mobile Radio Communications 2003, Vol. 1, pp.187-191, Sep. 2003. [11] A. Hasswa, N. Nasser, and H. Hossanein, Generic vertical handoff decision function for heterogeneous wireless, in Proc. of 2 nd IFIP Int. Conf. on Wireless and Optical Communications Networks, WOCN 2005, pp. 239-243, 6-8 March 2005. [12] H. Holma, and A. Toskala, WCDMA for UMTS, John Wiley, New York, 2004. [13] S. Toumpis, and A. J. Goldsmith, Capacity regions for wireless ad hoc networks, IEEE Transactions on Wireless Communications, Vol. 2, no. 4, July 2003. [14] IEEE Standard for Information technology Telecommunications and information exchange between systems. Local and metropolitan area networks. Specific requirements. Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications. [15] J. Laiho, A. Wacker, and T. Novosad, Radio Network Planning and Optimisation for UMTS, 2nd Edition, Chapter 6, pp. 333-337, Dec. 2005.