An improved WLAN-first access scheme for UMTS ... - Semantic Scholar
An Improved WLAN-first Access Scheme for UMTS/WLAN Integrated System Sibaram Khara
Iti Saha Mishra
Debashis Saha
School of Electronics Engineering VIT University Vellore, T.N. India +91 9443878634
Dept. of ETCE Jadavpur University Kolkata, W.B. India +91 9433083799
MIS & CS Group Indian Institute of Management Kolkata, W.B. India +91 3324622833
[email protected]
[email protected]
[email protected]
a request (i.e., call) is an important quality of service (QoS) metric. It depends upon the way a request is handled in the dual coverage of UMTS and WLAN. Therefore, interest is growing to design efficient technique for sharing of requests between UMTS and WLAN to improve the system capacity as well as QoS parameters in a mixed cell [2].
ABSTRACT In a tunnel-WLAN model of a UMTS/WLAN interworking system, a user always prefers to access the WLAN as soon as he/she moves to a WLAN hotspot. Thus a user never misses the higher bit rate of WLAN as long as its bandwidth is available. We have examined that overall dropping probability of a request in a mixed cell (i.e., a UMTS cell with underlying WLANs) improves if and only if the blocking probability of a request in WLAN remains lower than that in the UMTS system. Thus dropping probability increases with increasing blocking probability in WLAN. We propose a WLAN-first access scheme which transfers all blocked requests of WLAN to overlaying UMTS system, thereby preventing compulsory dropping of blocked requests in WLAN. This technique improves the request dropping probability in an entire mixed cell.
In a WLAN hotspot, requests can be handled broadly by two different schemes. Firstly, all new requests (NRs) (i.e., fresh requests) of mixed users are forced to access UMTS, and subsequently, blocked requests are transferred to WLAN. This is called UMTS-first scheme [3], [4]. In this model, users sometimes miss the higher bit rates of WLAN. The model provides low dropping probability of request at the cost of permanently withdrawing some WLAN spectrum to handle blocked requests of UMTS [3]. So, the model cannot be implemented using conventional standards of WLAN since it requires separate air interface on withdrawn WLAN spectrum. Secondly, NRs of mixed users are forced to access WLAN first. A mixed user always initiates vertical handoff request (VHR) from UMTS to WLAN (called downward VHR) as soon as he/she enters a WLAN hotspot with ongoing session. This is called WLAN-first scheme [5], [6]. In this model, WLAN hotspot is like a tunnel in the sense that a mixed user in a hotspot is never visible to UMTS. Hence it is also called tunnel-WLAN (T-WLAN) model. So, blocked requests are necessarily dropped in a mixed cell. We examine that request dropping probability in a T-WLAN model improves if and only if the request blocking probability in WLAN is less than that in UMTS.
Categories and Subject Descriptors Category of this article is networking track. The article proposed an efficient technique for handling calls i.e., request in a mixed cell comprising UMTS and WLAN.
General Terms Performance
Keywords Blocking, request (i.e., call), dropping, handoff, UMTS, WLAN, WLAN hotspot.
1. INTRODUCTION
T-WLAN model never deprives a mixed user of WLAN service as long as WLAN spectrum is available. Thus T-WLAN model with improved performance would be a better solution for efficient resource sharing in a mixed cell. We propose a modified WLAN-first access scheme in which a mixed user accesses WLAN on priority. The blocked requests of WLAN are transferred to UMTS. Therefore, a blocked NR in WLAN is dropped if it is subsequently blocked in UMTS also. Hence the technique minimizes dropping probability of a blocked request of WLAN. As a result, overall dropping probability improves in a mixed cell.
Integration of 3rd generation networks (e.g. Universal Mobile Telecommunication System (UMTS)) and wireless local area network (WLAN) is considered as a solution to fulfill some of the service demands of next generation networks (i.e., fourth generation (4G) networks) on hotspot basis [1]. An underlying WLAN cell within a UMTS cell coverage is called a WLAN hotspot. A UMTS cell with underlying isolated WLAN hotspot(s) is called a mixed cell and a UMTS user having WLAN privileges is called a mixed user. In a mixed cell, the dropping probability of
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. SAC’10, March 22-26, 2010, Sierre, Switzerland. Copyright 2010 ACM 978-1-60558-638-0/10/03…$10.00.
2. Review of Related Works The simulation model [5] based on T-WLANs evaluates that the capacity of a mixed cell increases by nearly three times with 25% WLAN coverage. The light and medium loaded hotspots can absorb up to 50% of the load of a congested UMTS cell, and the capacity of a UMTS cell increases by 50%. A resource sharing
630
established in an NR-successful state. An NR-successful state may move to completion state when a session in completed or terminated. Or, it may move to HHR-arrival state when a mobile station (MS) initiates HHR in neighbor cell. Or, it may move to VHR-arrival state in WLAN when MS moves to WLAN hotspot and initiates VHR. Similarly, HHR-arrival and VHR-arrival states of UMTS will move to other states as shown in Figure 1. Similarly, in WLAN, the NR-arrival moves to either NR-blocked or NR-successful state. NR-successful moves to either completion or VHR-arrival state (in UMTS). When an NR is blocked in WLAN, an MS automatically transfer the request to UMTS without user’s intervention.
based call admission control scheme (CAC) in a T-WLAN-based cellular/WLAN integrated nework is effective to improve handoff dropping probability [2]. A handoff request can be transferred to cellular system, if it is blocked in WLAN. The call handling algorithm uses resource sharing policy for only handoff requests. Thus, it decreases blocking probability of a handoff request at the expense of increasing blocking probability of fresh request. The vertical handoffs from WLAN to UMTS need to be supported and the related traffic is to be included in the analytical model. The analytical approach proposed in [7] models the mobilitypatters in the 3G-WLAN integrated systems by correlating cell residence time (CRT) in a 3G cell with that in WLAN hotspots. Model is useful to estimate the horizontal and vertical handoff arrival rates. A user’s access is dependent on admission parameters for efficient grant of voice and data requests in the cellular and WLAN systems. Resource utilization can be maximized under proper load balancing between cellular and WLANs. However, this model requires some new signaling for handoff between cellular and WLAN for exchange of network’s information for admission decision.
VHR-arrival in WLAN moves to either VHR-blocked or VHRsuccessful state. A VHR-successful transits to either completion state or VHR-arrival state (in UMTS). When an MS moves to WLAN, before accessing WLAN, optionally, it can send a standby message to keep the UMTS-attach state alive for a period of time [1]. This enables an MS to continue UMTS session when it gets back to UMTS when VHR is denied in WLAN. Thus, an MS can continue its session in UMTS after blocking of VHR in WLAN. So, VHR-blocked state moves successful state (in UMTS) from which it had moved to VHR arrival in WLAN. Thus a VHRblocked state in WLAN is actually a successful state in UMTS which can directly move to completion or HHR-arrival state of UMTS.
3. Overview of Proposed System Model We describe a model of a mixed cell and mobility of users with the state diagram as shown in Figure 1. The big hexagon represents a UMTS cell, and the big ellipse inside it presents the equivalent WLAN of all WLAN hotspots.
Neighbor mixed cell
UMTS-only coverage
λunr NR-arrival
b1 Blocked
b2
λuhhr HHR-arrival
VHR-successful
VHR-arrival w λvhr
VHR-successful
w λvhr
w ( Pvhs )′
NR-Blocked
NR-successful
b
b
( Pnsw )′
b′
u g′( Pvhs )
u ′ u ′ u ′ ( Pns ) , ( Phhs ) , ( Pvhs ) u u u Pns , Phhs , Pvhs VHR-Blocked
Completion
NR-arrival
λhhr
HHR-arrival u
u ′ g ′( Pvhs )
g
g
b′
u g′( Phhs )
u ′ g ′( Phhs ) Completion
HHR-successful
w Pvhs
Pnsw
u ′ g ′( Pns )
g
b2′
VHR-arrival
1
NR-successful
b2′
λuvhr
b2
u g′( Pns )
b1′
λwnr
NR-successful HHR-successful VHR-successful In UMTS
Mixed cell
WLAN coverage
Figure 1: Flow graphs of states of requests and transition probabilities.
3.1 States of Requests and Flow Graphs
3.2 State Transition Probabilities Assume, blocking probabilities of NR, HHR (or VHR) in UMTS are b1 and b2 , respectively. Request blocking probability in
There are three events in UMTS-only coverage; NR, horizontal handoff request (HHR), and VHR from WLAN to UMTS (i.e., upward VHR). There are two events in WLAN hotspots; NR and downward VHR. The origins of NR, HHR and VHR in UMTSonly coverage are represented by NR-arrival, HHR-arrival and VHR-arrival states, respectively. The origins of NR and VHR in WLAN are represented by NR-arrival and VHR-arrival states (Fig. 1). If an NR is denied by the UMTS system, it moves from NR-arrival to blocked state. If an NR is granted a channel, it moves from NR-arrival to NR-successful state. A session is
WLAN is b . For any value x , we represent (1 − x ) by x ′ . In UMTS-only coverage, transition probability from NR-arrival
′
to blocked state is b1 and to NR-successful state is b1 . Assume g is the coverage fraction which is the ratio of the coverage of all WLAN hotpots to the coverage of a pure UMTS cell. All events in a mixed cell are shared by WLAN in proportion with g
631
[3]. So, NR-successful moves to VHR-arrival state in WLAN with probability g . Probability that NR-arrival does not move to VHRarrival in WLAN is g ′ . Probability that a new session (NS)
From Figure 1, probabilities that NR-arrival, HHR-arrival and VHR-arrival in UMTS generate VHR-arrival in WLAN are (1 − b1 ) g , (1 − b2 ) g and (1 − b2 ) g , respectively. VHR arrival
initiates HHR in neighbor cell is Pnsu . So, NR-arrival moves to
w rate λ vhr in WLAN is given by,
w λ vhr = λunr b1′ g + λuhhr b2 ′ g + λuvhr b2 ′ g
u HHR-arrival and completion states with probability g ′Pns and
(2)
u ′ u u g ′( Pns ) . Phhs and Pvhs are the probabilities that horizontally
Similarly, using Figure 1, we can write the VHR arrival rate in UMTS as follows.
handed over session (HHS) and vertically handed over (VHS) will initiate HHR in neighbor cell. These are provided in Section 6. In UMTS-only coverage, the transition probabilities from HHRarrival and VHR-arrival states to other states can be determined as shown in Figure 1.
w w λuvhr = λ wnr b ′Pnsw + λ vhr b ′Pvhs
Using equations (1), (2) and (3) we can solve for
,
λuvhr
and
u w λ vhr when λ nr is known to us.
In WLAN, NR-arrival state moves to NR-blocked and NRsuccessful states with probabilities b and b ′ , respectively. NRsuccessful transits to VHR-arrival state in UMTS and completion
5. Steady State Probability of Sessions We assume there are total M channels in a UMTS cell, out of which maximum m channels can be used for NRs and remaining ( M − m) channels are reserved for handoff requests. Using Erlang's loss-formula for three dimensional steady state Markov chain [8], the probability that there are i NR-successful, j HHR-
state in WLAN with probabilities Pnsw and ( Pnsw ) ′ , respectively.
Pnsw is the probability that an NS of WLAN initiates VHR in UMTS. All blocked NRs in WLAN are transferred as NRs in UMTS. So, transition probability from NR-blocked (in WLAN) to NR-arrival (in UMTS) is one. NR-arrival state in WLAN transits to VHR-successful and VHR-blocked states with probabilities b and b ′ , respectively. VHR-successful state moves to VHRarrival state in UMTS and completion states in WLAN with
successful and k VHR-successful states in the UMTS system of a mixed cell is given as follows.
(
)(
)(
) ⎫⎪⎬ (4)
j u ⎧ u u i u u u k ⎪ λnr E [Tns ] λhhr E[Thhs ] λvhr E[Tvhs ] P u (i , j , k ) = P u ( 0, 0,0 ) ⎨ i! j!k! ⎪ ⎩
w w w probabilities Pvhs and ( Pvhs is the ) ′ , respectively. Pvhs probability that an VHS in WLAN initiates again VHS in UMTS. When a VHR is blocked in WLAN, an MS resumes the ongoing session in UMTS. VHR-blocked state moves back to NRsuccessful, or HHR-successful, or VHR-successful state (of UMTS) when users resides in WLAN coverage. For simplicity these three states are shown by a single node. These states will transit to HHR-arrival with the probabilities u u u Pns , Phhs and Pvhs , respectively, which are represented by u u u Pns , Phhs , Pvhs . These states will transit to completion state
with probabilities
(3)
λuhhr
⎪ ⎭
P (0,0,0) is the probability that no request is being processed in a u u UMTS cell. T nsu , T hhs and Tvhs represent channel holding times (CHTs) of NS, HHS and VHS, respectively.
(
)
(
)
(
)
i j k ⎤ ⎡m u λ E[T u ] ⎧⎪M −i λu E[T u ] ⎛ M −i− j λu E[T u ] ⎞⎫⎪ Pu (0,0,0) = ⎢∑ nr ns ⎨ ∑ hhr hhs ⎜⎜ ∑ vhr vhs ⎟⎟⎬⎥ ⎢i=0 i! j! k! ⎜ k =0 ⎟⎪⎥ ⎪⎩ j =0 ⎝ ⎠⎭⎥⎦ ⎢⎣ Similarly, in WLAN with total K channels, we can easily estimate the probability of i NR-successful and k VHR-successful states. We skip this due to length limitation.
u ′ u ) ′ and ( Pu )′ , respectively, ( Pns ) , ( Phhs vhs
6. Mean Channel Holding Time
u u u which are represented by ( Pns )′, ( Phhs )′, ( Pvhs )′ .
We assume call (i.e., request) holding time (caht) has exponential distribution with mean 1 h sec in entire mixed cell. Cell residence
4. Handoff Estimation
time (crt) has hiper-Erlang distribution with mean 1 r u sec in
The transition probability from one state to any other state is given by the product of the probabilities of all transitions that are involved between the two states. From Figure 1, the probabilities that NR-arrival, HHR-arrival and VHR-arrival (in UMTS) generate HHR-arrival in neighbor cell are
UMTS-only coverage and 1 r w sec in WLAN. Say, X is the general random variable which represents CHT. Distribution of X is given by the distribution of the minimum of the distributions of crt and caht [3]. Probability distribution function F X (t ) and density function, f X (t ) are as follows.
u u (b1′ g ′Pns + b1′ gbPns ), (b ′ g ′P u + b ′ gbP u ) , 2 u hhr
λ
vhs
and
2 vhs u represent vhr
λ
u u (b2 ′ g ′Phhs + b2 ′ gbPhhs ),
respectively.
We
and
FX (t ) = P ( X ≤ t ) = 1 − P (caht > t , residual _ crt > t )
assume λunr ,
f X (t ) =
the NR, HHR and VHR arrival rates,
respectively, in UMTS. Writing the flow balance equation under equilibrium state (i.e., flow in=flow out across UMTS cells), we write the following. λu = λu b ′ (1 − gb′)P u + λu b ′ (1 − gb′) P u + λu b (1 − gb′)P u (1) hhr
nr 1
ns
hhr 2
hhs
vhr 2
α ⎞ d⎛ α ⎜1 − ∫ f caht (t )dt ∫ f residual _ crt (t )dt ⎟ ⎜ ⎟ dt ⎝ t t ⎠
Using Laplace transform approach [8], the mean value of X can *(1)
be obtained as follows. E[ X ] = (−1) f X the Laplace transform of
vhs
632
(0) , where f X* ( s ) is
f X (t ) and f X*(1) (0) is the first
−1
and that occurs in the WLAN system with probability g . A blocked NR in WLAN accesses UMTS, the dropping probability of a blocked NR of WLAN is gbb1 . So, net dropping probability,
derivative of f X* ( s ) at s = 0 . From [8], we directly write mean CHT for different sessions as follows.
(
(
)
)
ru 1 ru * u u * ] = E[Tvhs = 1 − f crt ( h) − 1 − f crt ( h) , E[Thhs h h h2 rw w * 1 rw w * [ ] = 1− f crt E T and E[Tns ]= − 1 − f crt ( h ) vhs − w (h) −w h h h2 u E[Tns ]=
(
(
)
Pdnr of an NR in an entire mixed cell is the sum of the dropping probabilities of NRs of both UMTS and WLAN.
)
Pdnr = (1 − g )b1 + gbb1 Or, b1 − Pdnr = gb1 (1 − b)
Equation (9) implies that the dropping probability of NR in a mixed cell without WLAN coverage (i.e., g = 0 ) is equal to
* * f crt (h) and f crt − w ( h) are Laplace transforms for probability
density functions of crt in UMTS and WLAN, respectively, at s = h . Laplacc transform of hiper-Erlang distribution is,
N ⎛ nθ * (s) = f crt ∑ α i ⎜⎜ i i i =1 ⎝ s + niθ i
⎞ ⎟ ⎟ ⎠
that in pure UMTS cell. In equation (9), if both g and b are positive fractions, then Pdnr < b1 . Therefore, a mixed cell with underlying WLAN coverage always yields lower dropping probability of NR. Now consider that blocked requests of WLAN are not transferred to UMTS [5]. In that case, the dropping probability of NR is given, b1 − Pdnr = g (b1 − b) (10) It is seen from equation (10) that dropping probability of an NR in a mixed cell is lower than that in a pure UMTS cell if and only if the blocking probability of an NR in WLAN is less than that in a pure UMTS cell i.e., only if b < b1 . Our access scheme transfers blocked requests of WLAN to UMTS, so the dropping probability of NR is less affected by increasing blocking probability of WLAN. Dropping Probability of HHR in UMTS: In UMTS-only coverage, dropping probability of an HHR Pdhhr can be written as follows. Pdhhr = (1 − g )b2 = g ′b2 (11)
ni (5)
N
Where α i ≥ 0 , ∑ α i = 1 and N , ni , θ i are positive numbers. i =0
u Pns
u u = P(caht > residual_ crt) and Phhs = Pvhs = P(caht > crt) . ∞∞
u Pns = ∫ ∫ f residual _ crt (t ) f caht (τ )dτdt =
(
* r u 1 − f crt ( h) h
)
0t u u * u u Similarly, Phhs = Pvhs = f crt ( h) = 1 − (1 r )(hPns ) * ( h) r w 1 − f crt w w w Pnsw = and Pvhs = 1 − (1 r )(hPns )
(
)
h
7. Performance Analysis 7.1 Blocking Probabilities in UMTS
In equation (11), Pdhhr < b2 for all positive values of g . So, dropping probability of an HHR in a mixed cell is always less than that in a pure UMTS cell. Dropping probability of VHR in UMTS: All blocked VHRs in UMTS are necessarily dropped. VHR-arrivals in UMTS occur when NSs and VHSs moves from WLAN to UMTS. From Figure 1, the probability of occurrence of VHR in UMTS due to moving
Blocking probability of NR in UMTS: An NR-arrival moves to blocked state when there are already m NR-successful states in i < m , ( j + k ) > ( M − m) and UMTS or when
(i + j + k ) = M . So b1 is given by the following. M −m M −m− j u
b1 = ∑
j =0
∑
k =0
m−1M −i
P (m, j, k ) + ∑ ∑ P(i, j, M − i − j) (6)
′
i=0 j =0
′
(
w
w
)
coverage is blocked with probability b2 . So, dropping probability u of VHR, Pdvhr is given by,
(7)
(
u w w Pdvhr = b2 gb ′ Pvhds + Pnds
)
(12)
Dropping Probability of VHR in WLAN: VHR occurs in WLAN with probability g . But all blocked VHRs in WLAN retain the UMTS sessions and are not dropped. Thus, dropping probability of a VHR in WLAN,
Blocking Probability in WLAN: An NR or a VHR is blocked when there are already K successful states in the system. K
∑ P w (i, K − i)
w
occurs in UMTS is gb Pvhds + Pnds . A VHR in UMTS-only
m m −i
So, b =
′
w is g (1 − b) Pvhds i.e., gb Pvhds . So, total probability that a VHR
handoff request (HHR or VHR) (i.e., b2 ) is given by,
i =0 j =0
w
of an NS from WLAN to UMTS is gb Pnds . Probability that a VHR occurs due to a VHS initiating VHR from WLAN to UMTS
Handoff blocking probability in UMTS: An HHR-arrival or a VHR-arrival moves to blocked state when there are already M successful states in UMTS. This situation arises when ( j + k ) ≥ ( M − m) and i + j + k = M . Blocking probability of
b2 = ∑ ∑ P u (i, j , M − i − j )
(9)
(8)
w Pdvhr =0
i =0
(13)
Dropping Probability of handoff request (HR) (i.e., HHR or VHR): Dropping probability of HR in a mixed cell, is sum of the dropping probabilities of HHRs and VHRs. Thus, from equations (11), (12) and (13), dropping probability of an HR, Phr in a mixed cell is as follows.
7.2 Dropping Probability in a Mixed Cell Dropping Probability of NR: In a mixed cell, an NR-arrival in UMTS-only coverage occurs with probability (1 − g ) i.e. g ′
633
u u w Phr = Pdhhr + Pdvhr + Pdvhr
[ (
w w Or, b2 − Phr = gb 2 1 − b ′ Pvhds + Pnds
)]
less than 9 (approximately) per second. Beyond the request rate of 10 per sec, the dropping probability of an NR in a mixed cell is more than that of a pure UMTS cell. Figure 2(b) shows that the blocking probability of WLAN exceeds that of the UMTS system, but dropping probability of NR in a mixed cell never exceeds that of the UMTS system. This is due to the fact that all blocked WLAN requests are transferred to UMTS, thereby decreasing the dropping probability. Figure 2(c) shows the change in dropping probability of NR in a mixed cell with increasing new traffic with the effect of g. As g increases this dropping probability decreases. At NR arrival rate of 14 per sec and g = 0.40 , the dropping probability decreases by 67.4% with respect to a pure UMTS cell (i.e., g = 0 ). To provide this performance level the UMTS-first access scheme [3] needs at least 10 reserved WLAN channels. Our scheme additionally permits NRs to access WLAN and supports session-handover from UMTS to WLAN, its request dropping performance is quite comparable with UMTS-first scheme without reserving WLAN bandwidth for handoff handling.
(14)
In equation (14), if g = 0 , then Phr = b2 . So, dropping probability of a HR in a mixed cell without WLAN coverage is equal to the dropping probability of a HHR in a pure UMTS cell.
(
w w If g , b ′ , b2 and Pvhds + Pnds
)
are positive fractions, then
Blocking/dropping probabilty
(b2 − Phr ) > 0 . Therefore, dropping probability of a HR in a mixed cell is always less than that in a pure UMTS cell. NR blocking in UMT S NR blocking in WLAN NR dropping in a mixed cell
0.4 0.3 0.2 0.1 0 0
Blocking /dropping probability
(a)
2 4 6 8 10 12 14 16 18 20 NR arrival rate from WLAN users
0
2
Dropping probability of NR in a mixed cell
4
6
8
Proposed WLAN-first access scheme improves the dropping probability of requests which is an important QoS metric in a mixed cell. The blocked requests of WLAN are necessarily not dropped since they are redirected to UMTS. The CAC also ensures that no ongoing session is dropped in WLAN when VHR is denied. The proposed CAC scheme avoids the effect of increasing WLAN traffic on the traffic of mixed users.
10 12 14 16 18 20
NR arrival rate from WLAN users
(b) 0.40
References
g=0 g=0.1 g=0.2 g=0.3 g=0.4
0.30 0.20
[1]
[2]
0.10 0.00 0
(c)
9. Conclusion
NR blocking in UMT S NR blocking in WLAN NR dropping in a mixed cell
0.5 0.4 0.3 0.2 0.1 0
2
4
[3]
6 8 10 12 14 16 18 20 NR arrival rate (per sec)
[4]
Figure 2: Blocking/dropping probability Versus WLAN traffic in (a) TWLAN model and (b) in proposed model, and (c) dropping probability of NR with increasing new traffic in a mixed cell.
[5]
8. Numerical Results We compare the request dropping performance between T-WLAN and proposed access schemes. We use n = 2 , n1 = 3 , n 2 = 2 ,
[6]
α 1 = 0.4 , α 2 = 0.6 , θ1 = 0.45 and θ 2 = 0.35 in equation (5)
[7]
and set h = 2.5 sec , r = 3.31 sec , M = 30 , m = 5 for UMTS system. For WLAN, we use n = 2 , n1 = 2 , n 2 = 2 , u
α 1 = 0.2 , α 2 = 0.8 , θ1 = 0.25 and θ 2 = 0.35 in equation (5),
[8]
and set r w = 2.75 sec and K = 50 for WLAN system. Request u arrival rate from mixed users λ nr = 14 per sec. WLAN traffic is increased by increasing the request arrival rate from WLAN users (i.e., users cannot access UMTS). Figure 2(a) shows that dropping probability of NR in a mixed cell remains low as long as blocking probability in the WLAN system is less than that of the UMTS system. This occurs till the request arrival rate of WLAN users is
634
S. -L. Tsao and C. -C. Lin, “Design and Evaluation of UMTS/WLAN Interworking Stragegies,” 56th IEEE Vehicular TechnologyConference (VTC’02), 2002-Fall., Vol. 2. pp. 777-781. E. Steven-Navarro and V. W. .S. Wong, “Resource Sharing in an Integrated Wireless Cellular/WLAN System,” Canadian Conference on Electrical and Computer Engineering, 2007 (CCECE 2007), Vancouver, Canada, April 2007, pp 631- 634. S. Tang and W. Li “Performance Analysis of the 3G Network with Complementary WLANs” Proceeding of IEEE Globecom 2005, pp 2637-2641. D. Chen, X. Wang and A. K. Elhakeem, “Performance Analysis of UMTS Handover with the Help of WLAN”, Proceeding of Qshine’05, IEEE 2005, Volume , Issue , Aug. 2005. H. Liu, H. Bhaskaran, D. Raychaudhuri and S Verma, “Capacity Analysis of a Cellular Data System with UMTS/WLAN Interworking” Proceeding of IEEE VTC 2003, vol 3, pp. 1817-1821. S. Khara, I. S. Misra and D. Saha, “Data Traffic Analysis in a UMTS Cell with Underlying Tunnel-WLANs,” Proc. of ICDCN 2009, Springer-Verlag, Vol. LNCS 5408, pp 389-394, 2009. S. Song, Y. Cheng, and W. Zhuang, “Improving Voice and Data Services in Cellular/WLAN Integrated Networks by Admission Control” IEEE Trans. on Wireless Comm., Vol. 6, no. 11, November 2007, pp 4025-4036. Y. Fang, “Hyper-Erlang Distribution Model and its Application in Wireless Mobile Networks,” Intl. Journal of Wireless Networks: Special issue: Design and modeling in mobile and wireless systsems, Vol 7, Issue 3, 2001, pp 211-219.
Iti Saha Mishra
Debashis Saha
School of Electronics Engineering VIT University Vellore, T.N. India +91 9443878634
Dept. of ETCE Jadavpur University Kolkata, W.B. India +91 9433083799
MIS & CS Group Indian Institute of Management Kolkata, W.B. India +91 3324622833
[email protected]
[email protected]
[email protected]
a request (i.e., call) is an important quality of service (QoS) metric. It depends upon the way a request is handled in the dual coverage of UMTS and WLAN. Therefore, interest is growing to design efficient technique for sharing of requests between UMTS and WLAN to improve the system capacity as well as QoS parameters in a mixed cell [2].
ABSTRACT In a tunnel-WLAN model of a UMTS/WLAN interworking system, a user always prefers to access the WLAN as soon as he/she moves to a WLAN hotspot. Thus a user never misses the higher bit rate of WLAN as long as its bandwidth is available. We have examined that overall dropping probability of a request in a mixed cell (i.e., a UMTS cell with underlying WLANs) improves if and only if the blocking probability of a request in WLAN remains lower than that in the UMTS system. Thus dropping probability increases with increasing blocking probability in WLAN. We propose a WLAN-first access scheme which transfers all blocked requests of WLAN to overlaying UMTS system, thereby preventing compulsory dropping of blocked requests in WLAN. This technique improves the request dropping probability in an entire mixed cell.
In a WLAN hotspot, requests can be handled broadly by two different schemes. Firstly, all new requests (NRs) (i.e., fresh requests) of mixed users are forced to access UMTS, and subsequently, blocked requests are transferred to WLAN. This is called UMTS-first scheme [3], [4]. In this model, users sometimes miss the higher bit rates of WLAN. The model provides low dropping probability of request at the cost of permanently withdrawing some WLAN spectrum to handle blocked requests of UMTS [3]. So, the model cannot be implemented using conventional standards of WLAN since it requires separate air interface on withdrawn WLAN spectrum. Secondly, NRs of mixed users are forced to access WLAN first. A mixed user always initiates vertical handoff request (VHR) from UMTS to WLAN (called downward VHR) as soon as he/she enters a WLAN hotspot with ongoing session. This is called WLAN-first scheme [5], [6]. In this model, WLAN hotspot is like a tunnel in the sense that a mixed user in a hotspot is never visible to UMTS. Hence it is also called tunnel-WLAN (T-WLAN) model. So, blocked requests are necessarily dropped in a mixed cell. We examine that request dropping probability in a T-WLAN model improves if and only if the request blocking probability in WLAN is less than that in UMTS.
Categories and Subject Descriptors Category of this article is networking track. The article proposed an efficient technique for handling calls i.e., request in a mixed cell comprising UMTS and WLAN.
General Terms Performance
Keywords Blocking, request (i.e., call), dropping, handoff, UMTS, WLAN, WLAN hotspot.
1. INTRODUCTION
T-WLAN model never deprives a mixed user of WLAN service as long as WLAN spectrum is available. Thus T-WLAN model with improved performance would be a better solution for efficient resource sharing in a mixed cell. We propose a modified WLAN-first access scheme in which a mixed user accesses WLAN on priority. The blocked requests of WLAN are transferred to UMTS. Therefore, a blocked NR in WLAN is dropped if it is subsequently blocked in UMTS also. Hence the technique minimizes dropping probability of a blocked request of WLAN. As a result, overall dropping probability improves in a mixed cell.
Integration of 3rd generation networks (e.g. Universal Mobile Telecommunication System (UMTS)) and wireless local area network (WLAN) is considered as a solution to fulfill some of the service demands of next generation networks (i.e., fourth generation (4G) networks) on hotspot basis [1]. An underlying WLAN cell within a UMTS cell coverage is called a WLAN hotspot. A UMTS cell with underlying isolated WLAN hotspot(s) is called a mixed cell and a UMTS user having WLAN privileges is called a mixed user. In a mixed cell, the dropping probability of
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. SAC’10, March 22-26, 2010, Sierre, Switzerland. Copyright 2010 ACM 978-1-60558-638-0/10/03…$10.00.
2. Review of Related Works The simulation model [5] based on T-WLANs evaluates that the capacity of a mixed cell increases by nearly three times with 25% WLAN coverage. The light and medium loaded hotspots can absorb up to 50% of the load of a congested UMTS cell, and the capacity of a UMTS cell increases by 50%. A resource sharing
630
established in an NR-successful state. An NR-successful state may move to completion state when a session in completed or terminated. Or, it may move to HHR-arrival state when a mobile station (MS) initiates HHR in neighbor cell. Or, it may move to VHR-arrival state in WLAN when MS moves to WLAN hotspot and initiates VHR. Similarly, HHR-arrival and VHR-arrival states of UMTS will move to other states as shown in Figure 1. Similarly, in WLAN, the NR-arrival moves to either NR-blocked or NR-successful state. NR-successful moves to either completion or VHR-arrival state (in UMTS). When an NR is blocked in WLAN, an MS automatically transfer the request to UMTS without user’s intervention.
based call admission control scheme (CAC) in a T-WLAN-based cellular/WLAN integrated nework is effective to improve handoff dropping probability [2]. A handoff request can be transferred to cellular system, if it is blocked in WLAN. The call handling algorithm uses resource sharing policy for only handoff requests. Thus, it decreases blocking probability of a handoff request at the expense of increasing blocking probability of fresh request. The vertical handoffs from WLAN to UMTS need to be supported and the related traffic is to be included in the analytical model. The analytical approach proposed in [7] models the mobilitypatters in the 3G-WLAN integrated systems by correlating cell residence time (CRT) in a 3G cell with that in WLAN hotspots. Model is useful to estimate the horizontal and vertical handoff arrival rates. A user’s access is dependent on admission parameters for efficient grant of voice and data requests in the cellular and WLAN systems. Resource utilization can be maximized under proper load balancing between cellular and WLANs. However, this model requires some new signaling for handoff between cellular and WLAN for exchange of network’s information for admission decision.
VHR-arrival in WLAN moves to either VHR-blocked or VHRsuccessful state. A VHR-successful transits to either completion state or VHR-arrival state (in UMTS). When an MS moves to WLAN, before accessing WLAN, optionally, it can send a standby message to keep the UMTS-attach state alive for a period of time [1]. This enables an MS to continue UMTS session when it gets back to UMTS when VHR is denied in WLAN. Thus, an MS can continue its session in UMTS after blocking of VHR in WLAN. So, VHR-blocked state moves successful state (in UMTS) from which it had moved to VHR arrival in WLAN. Thus a VHRblocked state in WLAN is actually a successful state in UMTS which can directly move to completion or HHR-arrival state of UMTS.
3. Overview of Proposed System Model We describe a model of a mixed cell and mobility of users with the state diagram as shown in Figure 1. The big hexagon represents a UMTS cell, and the big ellipse inside it presents the equivalent WLAN of all WLAN hotspots.
Neighbor mixed cell
UMTS-only coverage
λunr NR-arrival
b1 Blocked
b2
λuhhr HHR-arrival
VHR-successful
VHR-arrival w λvhr
VHR-successful
w λvhr
w ( Pvhs )′
NR-Blocked
NR-successful
b
b
( Pnsw )′
b′
u g′( Pvhs )
u ′ u ′ u ′ ( Pns ) , ( Phhs ) , ( Pvhs ) u u u Pns , Phhs , Pvhs VHR-Blocked
Completion
NR-arrival
λhhr
HHR-arrival u
u ′ g ′( Pvhs )
g
g
b′
u g′( Phhs )
u ′ g ′( Phhs ) Completion
HHR-successful
w Pvhs
Pnsw
u ′ g ′( Pns )
g
b2′
VHR-arrival
1
NR-successful
b2′
λuvhr
b2
u g′( Pns )
b1′
λwnr
NR-successful HHR-successful VHR-successful In UMTS
Mixed cell
WLAN coverage
Figure 1: Flow graphs of states of requests and transition probabilities.
3.1 States of Requests and Flow Graphs
3.2 State Transition Probabilities Assume, blocking probabilities of NR, HHR (or VHR) in UMTS are b1 and b2 , respectively. Request blocking probability in
There are three events in UMTS-only coverage; NR, horizontal handoff request (HHR), and VHR from WLAN to UMTS (i.e., upward VHR). There are two events in WLAN hotspots; NR and downward VHR. The origins of NR, HHR and VHR in UMTSonly coverage are represented by NR-arrival, HHR-arrival and VHR-arrival states, respectively. The origins of NR and VHR in WLAN are represented by NR-arrival and VHR-arrival states (Fig. 1). If an NR is denied by the UMTS system, it moves from NR-arrival to blocked state. If an NR is granted a channel, it moves from NR-arrival to NR-successful state. A session is
WLAN is b . For any value x , we represent (1 − x ) by x ′ . In UMTS-only coverage, transition probability from NR-arrival
′
to blocked state is b1 and to NR-successful state is b1 . Assume g is the coverage fraction which is the ratio of the coverage of all WLAN hotpots to the coverage of a pure UMTS cell. All events in a mixed cell are shared by WLAN in proportion with g
631
[3]. So, NR-successful moves to VHR-arrival state in WLAN with probability g . Probability that NR-arrival does not move to VHRarrival in WLAN is g ′ . Probability that a new session (NS)
From Figure 1, probabilities that NR-arrival, HHR-arrival and VHR-arrival in UMTS generate VHR-arrival in WLAN are (1 − b1 ) g , (1 − b2 ) g and (1 − b2 ) g , respectively. VHR arrival
initiates HHR in neighbor cell is Pnsu . So, NR-arrival moves to
w rate λ vhr in WLAN is given by,
w λ vhr = λunr b1′ g + λuhhr b2 ′ g + λuvhr b2 ′ g
u HHR-arrival and completion states with probability g ′Pns and
(2)
u ′ u u g ′( Pns ) . Phhs and Pvhs are the probabilities that horizontally
Similarly, using Figure 1, we can write the VHR arrival rate in UMTS as follows.
handed over session (HHS) and vertically handed over (VHS) will initiate HHR in neighbor cell. These are provided in Section 6. In UMTS-only coverage, the transition probabilities from HHRarrival and VHR-arrival states to other states can be determined as shown in Figure 1.
w w λuvhr = λ wnr b ′Pnsw + λ vhr b ′Pvhs
Using equations (1), (2) and (3) we can solve for
,
λuvhr
and
u w λ vhr when λ nr is known to us.
In WLAN, NR-arrival state moves to NR-blocked and NRsuccessful states with probabilities b and b ′ , respectively. NRsuccessful transits to VHR-arrival state in UMTS and completion
5. Steady State Probability of Sessions We assume there are total M channels in a UMTS cell, out of which maximum m channels can be used for NRs and remaining ( M − m) channels are reserved for handoff requests. Using Erlang's loss-formula for three dimensional steady state Markov chain [8], the probability that there are i NR-successful, j HHR-
state in WLAN with probabilities Pnsw and ( Pnsw ) ′ , respectively.
Pnsw is the probability that an NS of WLAN initiates VHR in UMTS. All blocked NRs in WLAN are transferred as NRs in UMTS. So, transition probability from NR-blocked (in WLAN) to NR-arrival (in UMTS) is one. NR-arrival state in WLAN transits to VHR-successful and VHR-blocked states with probabilities b and b ′ , respectively. VHR-successful state moves to VHRarrival state in UMTS and completion states in WLAN with
successful and k VHR-successful states in the UMTS system of a mixed cell is given as follows.
(
)(
)(
) ⎫⎪⎬ (4)
j u ⎧ u u i u u u k ⎪ λnr E [Tns ] λhhr E[Thhs ] λvhr E[Tvhs ] P u (i , j , k ) = P u ( 0, 0,0 ) ⎨ i! j!k! ⎪ ⎩
w w w probabilities Pvhs and ( Pvhs is the ) ′ , respectively. Pvhs probability that an VHS in WLAN initiates again VHS in UMTS. When a VHR is blocked in WLAN, an MS resumes the ongoing session in UMTS. VHR-blocked state moves back to NRsuccessful, or HHR-successful, or VHR-successful state (of UMTS) when users resides in WLAN coverage. For simplicity these three states are shown by a single node. These states will transit to HHR-arrival with the probabilities u u u Pns , Phhs and Pvhs , respectively, which are represented by u u u Pns , Phhs , Pvhs . These states will transit to completion state
with probabilities
(3)
λuhhr
⎪ ⎭
P (0,0,0) is the probability that no request is being processed in a u u UMTS cell. T nsu , T hhs and Tvhs represent channel holding times (CHTs) of NS, HHS and VHS, respectively.
(
)
(
)
(
)
i j k ⎤ ⎡m u λ E[T u ] ⎧⎪M −i λu E[T u ] ⎛ M −i− j λu E[T u ] ⎞⎫⎪ Pu (0,0,0) = ⎢∑ nr ns ⎨ ∑ hhr hhs ⎜⎜ ∑ vhr vhs ⎟⎟⎬⎥ ⎢i=0 i! j! k! ⎜ k =0 ⎟⎪⎥ ⎪⎩ j =0 ⎝ ⎠⎭⎥⎦ ⎢⎣ Similarly, in WLAN with total K channels, we can easily estimate the probability of i NR-successful and k VHR-successful states. We skip this due to length limitation.
u ′ u ) ′ and ( Pu )′ , respectively, ( Pns ) , ( Phhs vhs
6. Mean Channel Holding Time
u u u which are represented by ( Pns )′, ( Phhs )′, ( Pvhs )′ .
We assume call (i.e., request) holding time (caht) has exponential distribution with mean 1 h sec in entire mixed cell. Cell residence
4. Handoff Estimation
time (crt) has hiper-Erlang distribution with mean 1 r u sec in
The transition probability from one state to any other state is given by the product of the probabilities of all transitions that are involved between the two states. From Figure 1, the probabilities that NR-arrival, HHR-arrival and VHR-arrival (in UMTS) generate HHR-arrival in neighbor cell are
UMTS-only coverage and 1 r w sec in WLAN. Say, X is the general random variable which represents CHT. Distribution of X is given by the distribution of the minimum of the distributions of crt and caht [3]. Probability distribution function F X (t ) and density function, f X (t ) are as follows.
u u (b1′ g ′Pns + b1′ gbPns ), (b ′ g ′P u + b ′ gbP u ) , 2 u hhr
λ
vhs
and
2 vhs u represent vhr
λ
u u (b2 ′ g ′Phhs + b2 ′ gbPhhs ),
respectively.
We
and
FX (t ) = P ( X ≤ t ) = 1 − P (caht > t , residual _ crt > t )
assume λunr ,
f X (t ) =
the NR, HHR and VHR arrival rates,
respectively, in UMTS. Writing the flow balance equation under equilibrium state (i.e., flow in=flow out across UMTS cells), we write the following. λu = λu b ′ (1 − gb′)P u + λu b ′ (1 − gb′) P u + λu b (1 − gb′)P u (1) hhr
nr 1
ns
hhr 2
hhs
vhr 2
α ⎞ d⎛ α ⎜1 − ∫ f caht (t )dt ∫ f residual _ crt (t )dt ⎟ ⎜ ⎟ dt ⎝ t t ⎠
Using Laplace transform approach [8], the mean value of X can *(1)
be obtained as follows. E[ X ] = (−1) f X the Laplace transform of
vhs
632
(0) , where f X* ( s ) is
f X (t ) and f X*(1) (0) is the first
−1
and that occurs in the WLAN system with probability g . A blocked NR in WLAN accesses UMTS, the dropping probability of a blocked NR of WLAN is gbb1 . So, net dropping probability,
derivative of f X* ( s ) at s = 0 . From [8], we directly write mean CHT for different sessions as follows.
(
(
)
)
ru 1 ru * u u * ] = E[Tvhs = 1 − f crt ( h) − 1 − f crt ( h) , E[Thhs h h h2 rw w * 1 rw w * [ ] = 1− f crt E T and E[Tns ]= − 1 − f crt ( h ) vhs − w (h) −w h h h2 u E[Tns ]=
(
(
)
Pdnr of an NR in an entire mixed cell is the sum of the dropping probabilities of NRs of both UMTS and WLAN.
)
Pdnr = (1 − g )b1 + gbb1 Or, b1 − Pdnr = gb1 (1 − b)
Equation (9) implies that the dropping probability of NR in a mixed cell without WLAN coverage (i.e., g = 0 ) is equal to
* * f crt (h) and f crt − w ( h) are Laplace transforms for probability
density functions of crt in UMTS and WLAN, respectively, at s = h . Laplacc transform of hiper-Erlang distribution is,
N ⎛ nθ * (s) = f crt ∑ α i ⎜⎜ i i i =1 ⎝ s + niθ i
⎞ ⎟ ⎟ ⎠
that in pure UMTS cell. In equation (9), if both g and b are positive fractions, then Pdnr < b1 . Therefore, a mixed cell with underlying WLAN coverage always yields lower dropping probability of NR. Now consider that blocked requests of WLAN are not transferred to UMTS [5]. In that case, the dropping probability of NR is given, b1 − Pdnr = g (b1 − b) (10) It is seen from equation (10) that dropping probability of an NR in a mixed cell is lower than that in a pure UMTS cell if and only if the blocking probability of an NR in WLAN is less than that in a pure UMTS cell i.e., only if b < b1 . Our access scheme transfers blocked requests of WLAN to UMTS, so the dropping probability of NR is less affected by increasing blocking probability of WLAN. Dropping Probability of HHR in UMTS: In UMTS-only coverage, dropping probability of an HHR Pdhhr can be written as follows. Pdhhr = (1 − g )b2 = g ′b2 (11)
ni (5)
N
Where α i ≥ 0 , ∑ α i = 1 and N , ni , θ i are positive numbers. i =0
u Pns
u u = P(caht > residual_ crt) and Phhs = Pvhs = P(caht > crt) . ∞∞
u Pns = ∫ ∫ f residual _ crt (t ) f caht (τ )dτdt =
(
* r u 1 − f crt ( h) h
)
0t u u * u u Similarly, Phhs = Pvhs = f crt ( h) = 1 − (1 r )(hPns ) * ( h) r w 1 − f crt w w w Pnsw = and Pvhs = 1 − (1 r )(hPns )
(
)
h
7. Performance Analysis 7.1 Blocking Probabilities in UMTS
In equation (11), Pdhhr < b2 for all positive values of g . So, dropping probability of an HHR in a mixed cell is always less than that in a pure UMTS cell. Dropping probability of VHR in UMTS: All blocked VHRs in UMTS are necessarily dropped. VHR-arrivals in UMTS occur when NSs and VHSs moves from WLAN to UMTS. From Figure 1, the probability of occurrence of VHR in UMTS due to moving
Blocking probability of NR in UMTS: An NR-arrival moves to blocked state when there are already m NR-successful states in i < m , ( j + k ) > ( M − m) and UMTS or when
(i + j + k ) = M . So b1 is given by the following. M −m M −m− j u
b1 = ∑
j =0
∑
k =0
m−1M −i
P (m, j, k ) + ∑ ∑ P(i, j, M − i − j) (6)
′
i=0 j =0
′
(
w
w
)
coverage is blocked with probability b2 . So, dropping probability u of VHR, Pdvhr is given by,
(7)
(
u w w Pdvhr = b2 gb ′ Pvhds + Pnds
)
(12)
Dropping Probability of VHR in WLAN: VHR occurs in WLAN with probability g . But all blocked VHRs in WLAN retain the UMTS sessions and are not dropped. Thus, dropping probability of a VHR in WLAN,
Blocking Probability in WLAN: An NR or a VHR is blocked when there are already K successful states in the system. K
∑ P w (i, K − i)
w
occurs in UMTS is gb Pvhds + Pnds . A VHR in UMTS-only
m m −i
So, b =
′
w is g (1 − b) Pvhds i.e., gb Pvhds . So, total probability that a VHR
handoff request (HHR or VHR) (i.e., b2 ) is given by,
i =0 j =0
w
of an NS from WLAN to UMTS is gb Pnds . Probability that a VHR occurs due to a VHS initiating VHR from WLAN to UMTS
Handoff blocking probability in UMTS: An HHR-arrival or a VHR-arrival moves to blocked state when there are already M successful states in UMTS. This situation arises when ( j + k ) ≥ ( M − m) and i + j + k = M . Blocking probability of
b2 = ∑ ∑ P u (i, j , M − i − j )
(9)
(8)
w Pdvhr =0
i =0
(13)
Dropping Probability of handoff request (HR) (i.e., HHR or VHR): Dropping probability of HR in a mixed cell, is sum of the dropping probabilities of HHRs and VHRs. Thus, from equations (11), (12) and (13), dropping probability of an HR, Phr in a mixed cell is as follows.
7.2 Dropping Probability in a Mixed Cell Dropping Probability of NR: In a mixed cell, an NR-arrival in UMTS-only coverage occurs with probability (1 − g ) i.e. g ′
633
u u w Phr = Pdhhr + Pdvhr + Pdvhr
[ (
w w Or, b2 − Phr = gb 2 1 − b ′ Pvhds + Pnds
)]
less than 9 (approximately) per second. Beyond the request rate of 10 per sec, the dropping probability of an NR in a mixed cell is more than that of a pure UMTS cell. Figure 2(b) shows that the blocking probability of WLAN exceeds that of the UMTS system, but dropping probability of NR in a mixed cell never exceeds that of the UMTS system. This is due to the fact that all blocked WLAN requests are transferred to UMTS, thereby decreasing the dropping probability. Figure 2(c) shows the change in dropping probability of NR in a mixed cell with increasing new traffic with the effect of g. As g increases this dropping probability decreases. At NR arrival rate of 14 per sec and g = 0.40 , the dropping probability decreases by 67.4% with respect to a pure UMTS cell (i.e., g = 0 ). To provide this performance level the UMTS-first access scheme [3] needs at least 10 reserved WLAN channels. Our scheme additionally permits NRs to access WLAN and supports session-handover from UMTS to WLAN, its request dropping performance is quite comparable with UMTS-first scheme without reserving WLAN bandwidth for handoff handling.
(14)
In equation (14), if g = 0 , then Phr = b2 . So, dropping probability of a HR in a mixed cell without WLAN coverage is equal to the dropping probability of a HHR in a pure UMTS cell.
(
w w If g , b ′ , b2 and Pvhds + Pnds
)
are positive fractions, then
Blocking/dropping probabilty
(b2 − Phr ) > 0 . Therefore, dropping probability of a HR in a mixed cell is always less than that in a pure UMTS cell. NR blocking in UMT S NR blocking in WLAN NR dropping in a mixed cell
0.4 0.3 0.2 0.1 0 0
Blocking /dropping probability
(a)
2 4 6 8 10 12 14 16 18 20 NR arrival rate from WLAN users
0
2
Dropping probability of NR in a mixed cell
4
6
8
Proposed WLAN-first access scheme improves the dropping probability of requests which is an important QoS metric in a mixed cell. The blocked requests of WLAN are necessarily not dropped since they are redirected to UMTS. The CAC also ensures that no ongoing session is dropped in WLAN when VHR is denied. The proposed CAC scheme avoids the effect of increasing WLAN traffic on the traffic of mixed users.
10 12 14 16 18 20
NR arrival rate from WLAN users
(b) 0.40
References
g=0 g=0.1 g=0.2 g=0.3 g=0.4
0.30 0.20
[1]
[2]
0.10 0.00 0
(c)
9. Conclusion
NR blocking in UMT S NR blocking in WLAN NR dropping in a mixed cell
0.5 0.4 0.3 0.2 0.1 0
2
4
[3]
6 8 10 12 14 16 18 20 NR arrival rate (per sec)
[4]
Figure 2: Blocking/dropping probability Versus WLAN traffic in (a) TWLAN model and (b) in proposed model, and (c) dropping probability of NR with increasing new traffic in a mixed cell.
[5]
8. Numerical Results We compare the request dropping performance between T-WLAN and proposed access schemes. We use n = 2 , n1 = 3 , n 2 = 2 ,
[6]
α 1 = 0.4 , α 2 = 0.6 , θ1 = 0.45 and θ 2 = 0.35 in equation (5)
[7]
and set h = 2.5 sec , r = 3.31 sec , M = 30 , m = 5 for UMTS system. For WLAN, we use n = 2 , n1 = 2 , n 2 = 2 , u
α 1 = 0.2 , α 2 = 0.8 , θ1 = 0.25 and θ 2 = 0.35 in equation (5),
[8]
and set r w = 2.75 sec and K = 50 for WLAN system. Request u arrival rate from mixed users λ nr = 14 per sec. WLAN traffic is increased by increasing the request arrival rate from WLAN users (i.e., users cannot access UMTS). Figure 2(a) shows that dropping probability of NR in a mixed cell remains low as long as blocking probability in the WLAN system is less than that of the UMTS system. This occurs till the request arrival rate of WLAN users is
634
S. -L. Tsao and C. -C. Lin, “Design and Evaluation of UMTS/WLAN Interworking Stragegies,” 56th IEEE Vehicular TechnologyConference (VTC’02), 2002-Fall., Vol. 2. pp. 777-781. E. Steven-Navarro and V. W. .S. Wong, “Resource Sharing in an Integrated Wireless Cellular/WLAN System,” Canadian Conference on Electrical and Computer Engineering, 2007 (CCECE 2007), Vancouver, Canada, April 2007, pp 631- 634. S. Tang and W. Li “Performance Analysis of the 3G Network with Complementary WLANs” Proceeding of IEEE Globecom 2005, pp 2637-2641. D. Chen, X. Wang and A. K. Elhakeem, “Performance Analysis of UMTS Handover with the Help of WLAN”, Proceeding of Qshine’05, IEEE 2005, Volume , Issue , Aug. 2005. H. Liu, H. Bhaskaran, D. Raychaudhuri and S Verma, “Capacity Analysis of a Cellular Data System with UMTS/WLAN Interworking” Proceeding of IEEE VTC 2003, vol 3, pp. 1817-1821. S. Khara, I. S. Misra and D. Saha, “Data Traffic Analysis in a UMTS Cell with Underlying Tunnel-WLANs,” Proc. of ICDCN 2009, Springer-Verlag, Vol. LNCS 5408, pp 389-394, 2009. S. Song, Y. Cheng, and W. Zhuang, “Improving Voice and Data Services in Cellular/WLAN Integrated Networks by Admission Control” IEEE Trans. on Wireless Comm., Vol. 6, no. 11, November 2007, pp 4025-4036. Y. Fang, “Hyper-Erlang Distribution Model and its Application in Wireless Mobile Networks,” Intl. Journal of Wireless Networks: Special issue: Design and modeling in mobile and wireless systsems, Vol 7, Issue 3, 2001, pp 211-219.
