J Supercomput (2012) 59:658–675 DOI 10.1007/s11227-010-0459-2

Optimized fast handover scheme in Mobile IPv6 networks to support mobile users for cloud computing Seonggeun Ryu · Kyunghye Lee · Youngsong Mun

Published online: 12 June 2010 © Springer Science+Business Media, LLC 2010

Abstract In the future cloud computing, users will heavily use mobile devices. Mobile networks for cloud computing should be managed efficiently as well as support seamless services to mobile users regardless of their locations and movements. Hence, in mobile networks for cloud computing, it is important to support seamless mobility management to mobile users who request real-time services such as VoIP, streaming, and interactive game playing. To support seamless mobility management for various wireless technologies in cloud computing, Mobile IPv6 (MIPv6) and fast handovers for MIPv6 (FMIPv6) have been studied. FMIPv6 has been emerged to reduce long handover latency and packet loss in MIPv6. FMIPv6 may provide seamless handover by minimizing the handover latency, and prevent packet loss through buffering and tunneling. FMIPv6 uses anticipation based on layer 2 trigger information, and consists of two operation modes such as the predictive mode and the reactive mode. Several works have been done to evaluate the performance of FMIPv6 in different network environments. However, the previous works did not consider the probability of predictive mode failure (PPMF) that distinguishes two operation modes. Even in the most previous work, two operation modes of FMIPv6 are evaluated separately. However, to accurately analyze the overall performance of FMIPv6, two operation modes should be analyzed altogether. In this paper, FMIPv6 combining two operation modes is analyzed considering the PPMF that is affected by the radius of a cell, velocity of mobile nodes, and the layer 2 triggering time. The effect of system parameters, such as the PPMF, the time required to process additional layer 3 signaling, and the layer 2 trigger time, is analytically investigated with respect to the signaling S. Ryu () · K. Lee · Y. Mun School of Computing, Soongsil University, Sangdo-dong, Dongjak-gu, Seoul, Korea e-mail: [email protected] K. Lee e-mail: [email protected] Y. Mun e-mail: [email protected]

Optimized fast handover scheme in Mobile IPv6 networks to support

659

cost and the packet delivery cost. Analytical results show a trade-off between performance and system parameters. Then we show methods to optimize overhead of FMIPv6. Finally, mobile networks for cloud computing can be efficiently managed through the optimized FMIPv6. Keywords Cloud computing · Mobility protocol · FMIPv6 · Performance analysis · Probability of predictive mode failure

1 Introduction Cloud computing is a recent trend in IT which moves computing and data away from desktop and portable devices into large data centers. Cloud computing becomes more and more popular in large-scale computing and data store, since it enables the sharing of computing resources which are distributed all over the world [1]. The one among the key driving forces behind cloud computing is the ubiquity of broadband and wireless networking [2]. Wireless networks provide a flexible data communication system which can extend the IP core networks to provide location-independent access [3]. Mobile users use Internet services over lightweight mobile devices rather than the traditional desktop PC, and they frequently request real-time services, such as VoIP, video, and audio streaming. Thus, mobile networks for cloud computing should provide real-time services to mobile users seamlessly. However, lots of network resources are needed to transfer traffic for real-time services. Furthermore, since the number of mobile users continuously increases, the aggregate of resources that they consume cannot be considered negligible [4, 5]. Therefore, efficient mobility management is important in mobile networks for cloud computing. In cloud computing, there are lots of wireless technologies to support mobile users, such as the 3G, 4G, long term evolution (LTE), WiMAX, and WiFi [6]. To support real-time services for various wireless technologies seamlessly, Mobile IPv6 (MIPv6) [8] is a representative protocol of mobility managements [7]. When a mobile node (MN) moves between different subnets, MIPv6 allows the MN to maintain connectivity through the handover process which consists of the link switching, movement detection, new address configuration, and the location update. During the handover process, there is a period that the MN is unable to send or receive packets, i.e., the handover latency. In MIPv6, long handover latency is a well-known problem, especially when a home agent (HA) or a correspondent node (CN) is located far away from a MN. Long handover latency causes high packet loss rate causing deterioration of real-time traffic. Moreover, if packets are lost during the handover, network resource may be wasted, since packets should be retransmitted. Thus, fast handovers for Mobile IPv6 (FMIPv6) [9] has been studied to reduce the handover latency and packet loss during the handover. If seamless fast handover is not supported, a lot of advanced applications of mobile cloud computing cannot be realized. For example, real-time social web services such as “Google Buzz” and “buzzed on Virgin Mobile” are now emerging as advanced social network services. These services are based on user’s location information and require mobile networks for cloud computing to provide seamless real-time services. Since in mobile networks for cloud computing, realtime services will be heavily requested by advanced applications, features of FMIPv6

660

S. Ryu et al.

is greatly efficient due to its capability minimizing handover latency. That is, FMIPv6 will be a big asset for the real-time services, because it prevents the retransmission delay due to lost packets. Then networks for cloud computing will provide seamless mobility services for a lot of advanced location-based real-time applications. FMIPv6 provides seamless handover using anticipation based on layer 2 trigger information to reduce the handover latency and packet loss. FMIPv6 consists of two operation modes such as the predictive mode and the reactive mode. If the additional layer 3 signaling for FMIPv6 may be processed before the layer 2 handover is completed, FMIPv6 will be at the predictive mode. If the additional layer 3 signaling processing time is greater than the time between the layer 2 trigger and link down, FMIPv6 will switch to the reactive mode. In the case that the predictive mode is failed, we call this status as the predictive mode failure. In this paper, the PPMF is an abbreviation for the probability of predictive mode failure. Performance evaluations of various IPv6-based mobility protocols have been performed [10, 11]. However, they use the simulation and testbed approach for the performance evaluation. Several studies have been done to evaluate the performance of FMIPv6 in different network environments through mathematical analysis [12–15]. An enhanced FMIPv6 scheme where the registration delay might be reduced by performing the registration process in advance was proposed in [12]. However, only the predictive mode was analyzed to evaluate performance. Various IPv6-based mobility protocols were analytically evaluated and compared in terms of the handover delay in [13]. To optimize the handover delay, H. Fathi et al. proposed to use the adaptive retransmission timer. However, the predictive mode and the reactive mode were separately analyzed. Also, they only analyzed the handover latency. In [14], FMIPv6 is analyzed through the signaling cost, the packet delivery cost, and the total cost as a function of the layer 2 trigger. S. Pack and Y. Choi simply determined the success probability which means the probability of the predictive mode success using the layer 2 trigger time. However, FMIPv6 scheme that they used has been obsolete. There is a new standardized scheme that has to be analyzed. Also, they did not consider the probability of predictive mode success with network parameters. In [15], Makaya and Pierre compared various metrics of different IP mobility management schemes. They proposed analytical framework in order to perform a comprehensive analysis of the overall performance of various IPv6-based mobility protocols, such as MIPv6, FMIPv6, hierarchical MIPv6, and fast handover for hierarchical MIPv6. However, the packet delivery cost of FMIPv6 is calculated using fixed value for the prediction probability which has the opposite meaning to the PPMF. For the more accurate performance evaluation, the prediction probability should be obtained as the function of network parameters such as the radius of a cell, the time required to process additional layer 3 signaling, and the layer 2 triggering time. The previous works neither consider the PPMF, nor use various network parameters for the PPMF. Also, in the most previous papers, two operation modes of FMIPv6 were evaluated separately. It is essential to analyze the overall performance of a protocol by considering two operation modes altogether. Even in the researches considering two operation modes altogether, obsolete model of FMIPv6 was used or value of the PPMF was fixed. In this paper, FMIPv6 combining two operation modes is analyzed considering the PPMF that is affected by the radius of a cell, velocity of

Optimized fast handover scheme in Mobile IPv6 networks to support

661

mobile nodes, and the layer 2 triggering time. The effect of system parameters, such as the PPMF, the time required to process additional layer 3 signaling, and the layer 2 trigger time, is investigated for performance evaluation with respect to the signaling cost and the packet delivery cost. Numerical results show a trade-off between performance and system parameters. Then, methods to optimize FMIPv6 are shown in results through system parameters such as the radius of a cell, the time required to process additional layer 3 signaling, and the layer 2 triggering time. The rest of this paper is organized as follows. MIPv6 and FMIPv6 that are mobility management for mobile networks of cloud computing are presented in Sect. 2. Section 3 describes analytical model for performance evaluation. In Sect. 4, analytical results based on this analytical model is then investigated. Finally, we conclude the paper with some remarks and comments.

2 Fast handover for Mobile IPv6 Mobility management with provision of seamless handover is the key topic in mobile network for cloud computing. Then MIPv6 has been studied in the internet engineering task force (IETF) to provide seamless mobility and service continuity in intelligent and efficient ways. MIPv6 makes a MN stay connected to the Internet regardless of its location through the handover process. The handover process happens when the MN moves from one subnet to another, and it should accomplish four operations: link switching (i.e., the layer 2 handover), movement detection, new care-of address (CoA) configuration, and binding update (BU). During the handover period, the MN is unable to send or receive packets as usual. The length of this period which is called handover latency is very critical for the delay-sensitive and real-time services. In MIPv6, the MN has two addresses, such as a home address (HoA) and a CoA. The HoA is an address assigned to the MN used as the permanent address of the MN, and the CoA is an address associated with the MN while visiting a foreign subnet. MIPv6 allows the MN to move from one subnet to another without changing the MN’s HoA. The MN configures a CoA when it moves to a new subnet. Then the MN informs the home agent (HA) of the CoA through binding update process. Packets may be routed to the MN’s CoA using MN’s HoA regardless of the MN’s current point of attachment to the Internet. The MN may also continue to communicate with CNs after moving to the new subnet. In MIPv6, movement detection, new CoA configuration, and binding update are performed after the layer 2 handover is completed. FMIPv6 provides seamless handover using anticipation based on layer 2 trigger information to reduce the handover latency and packet loss in MIPv6. In FMIPv6, movement detection and new address configuration which are portions of the layer 3 handover are performed prior to the layer 2 handover. FMIPv6 consists of two operation modes such as the predictive mode and the reactive mode. If the additional layer 3 signaling for FMIPv6 may be processed before the layer 2 handover is completed, FMIPv6 will be at the predictive mode. If the additional layer 3 signaling processing time is greater than the time between the layer 2 trigger and link down, FMIPv6 will switch to the reactive mode. In FMIPv6, the predictive mode and the reactive mode are distinguished by the PPMF.

662

S. Ryu et al.

Fig. 1 Two operation modes of fast handovers for MIPv6

Figure 1(a) shows the message sequence of the predictive mode of FMIPv6. The MN sends a router solicitation for proxy (RtSolPr) message to a previous access router (PAR) to request information of the NAR. The PAR delivers information of the NAR to the MN by a proxy router advertisement (PrRtAdv) message. The MN configures a new care-of address (NCoA) used in the NAR. Then, the MN sends a fast binding update (FBU) message to the PAR to bind a previous care-of address (PCoA) with the NCoA. On receipt of the FBU message, the PAR attempts to establish a tunnel between itself and the NAR by exchanging a handover initiate (HI) and a handover acknowledge (HAck) messages. When the NCoA is verified at the NAR, the tunnel is established. Then, packets that arrive at the PCoA are forwarded to the NAR through the tunnel during the layer 2 handover. Upon receipt of the HAck message, the PAR sends a fast binding acknowledgement (FBack) messages to the MN and the NAR. The NAR buffers packets until it receives a fast neighbor advertisement (FNA) message from the MN. When the MN attaches to the NAR, it sends the FNA message to the NAR. Then the MN registers the NCoA to the HA. The reactive mode of FMIPv6 is shown in Fig. 1(b). The MN requests information of the NAR by sending the RtSolPr message. The FBU message encapsulated in the FNA message is sent to the PAR via the NAR after the layer 2 handover. After that, the MN registers the NCoA to the HA. In the reactive mode, the PAR may buffer packets to the MN before the layer 2 handover is begun. In this case, packets will be forwarded to the MN via the NAR when the PAR receives the FBU message. If the PAR chooses not to buffer packets, packets are lost during the handover. In this paper, the buffering of packets by the PAR during the handover is considered, since such buffering is useful in the reactive mode of FMIPv6 [9].

Optimized fast handover scheme in Mobile IPv6 networks to support

663

3 Analytical models In IPv6-based wireless networks, QoS may be defined by signaling overhead, handover latency, and packet loss [15]. Furthermore, in FMIPv6, the PPMF is an important parameter to evaluate FMIPv6 with MNs’ velocity, the radius of a cell, and the layer 2 trigger time. Analytical framework using the PPMF for evaluating performance of FMIPv6 is proposed in this section. We define that handover is begun from the layer 2 trigger epoch and is ended at which the MN receives the first packet in the new network. 3.1 Mobility modeling We assume that mobile service area are composed location areas (LAs) which are aggregate of cells as shown in Fig. 2(a) [16]. The innermost cell “0” is called the center cell; cells labeled “1” form the first ring around cell “0” and so forth. Each ring is labeled according to its distance from the center such that ring r1 refers to the cells in the first ring away from cell “0.” In general, rk (k = 1, 2, . . .) refers to the kth ring away from the center cell. The number of cells in kth ring is 6 · k. Then the number of cells N(K) is calculated as N (K) =

K 

6 · k + 1 = 3(K + 1) · K + 1,

(1)

k=1

where K denotes the outermost ring within the LA [17]. Given that the cell radius is r, we can observe that the perimeter of the center cell is 6r and the perimeter of the first ring is 18r. The radius r is determined based on the number of MNs and bandwidth allocation schemes. The perimeter L(K) and the

Fig. 2 Mobile service area

664

S. Ryu et al.

coverage area S(K) are L(K) = (12K + 6) · r   S(K) = 3(K + 1) · K + 1 · 2.6 · r 2 ,

(2)

where 2.6 · r 2 is the area of each cell [17]. We use fluid-flow model for mobility model, referring to [18] and [19]. Under the fluid-flow model in [17], the direction of an MN’s movement in the LA is uniformly distributed in the range of (0, 2π). Let v be average speed; S(K) and L(K) be area and perimeter of the LA, respectively. The average handover rate μh is equal to the average number of crossings of the boundary of the LA per unit time, then μh =

vL(K) . πS(K)

(3)

3.2 The probability of predictive mode failure (PPMF) In FMIPv6, additional signaling messages may be exchanged between an MN and an PAR, while the MN is in the overlapped area of cells in the PAR and the NAR. The predictive mode will be failed due to the MN leaving the overlapped area, before the additional signaling for FMIPv6 can be completed. Figure 2(b) shows the overlapped area of boundary cells. In this section, we now calculate the PPMF. Let T be a random variable for the time from layer 2 trigger epoch to link down (i.e., pending L2 handover). The PPMF, Prf , which the MN leaves the overlapped area before the required time for additional signaling in FMIPv6, Tfast , is  Pr(T < Tfast ), if Tfast ≤ Ttrigger Prf = (4) 1, if Tfast > Ttrigger , where Ttrigger is the time taken from the occurrence of the layer 2 trigger event to the start of the layer 2 handover. Tfast means the time duration till the tunnel between the PAR and the NAR is established since the layer 2 trigger occurred. In the case of Tfast > Ttrigger , FMIPv6 always operates in the reactive mode. In the other case, FMIPv6 may operate in ether the predictive mode or the reactive mode according to the PPMF. If we assume that T is exponentially distributed [20], then Prf becomes Prf = 1 − e−λTfast ,

(5)

where λ is the arrival rate of MNs into the overlapped area. For an MN whose direction of travel is uniform on the interval [0, 2π), we find that the arrival rate, λ, of the MN into the overlapped area is given by [21]: λ=

2vL , πS

(6)

where L is the length of the perimeter of the overlapped area, and S is the oversin(π/3) 2 1 2 lapped area. We can obtain L = 2 · ( 16 2πr) = 2πr r )= 3 and S = 2 · ( 6 πr − 2

Optimized fast handover scheme in Mobile IPv6 networks to support

665

r 2 (π−sin(π/3)) , referring to Fig. 2(b). MN’s velocity is 1000 3 3600

1 · 1000 · v, since in (5), units of length and time are the meter and the millisecond, respectively. Then

λ=

v . 900 · r(π − sin(π/3))

(7)

3.3 The total cost In this paper, we focus on costs occurred when an MN moves to other network, since FMIPv6 improves performance of MIPv6 handover. Thus, the costs are calculated by the number of signaling messages and delivered packets during the handover. The signaling cost is a cost induced by mobility management process, and the packet delivery cost is a sum of costs for data packets delivered during the handover. Therefore, the total cost, Ctotal , is considered as a sum of the signaling cost, Csignal , and the packet delivery cost, Cdelivery : Ctotal = α · Csignal + β · Cdelivery ,

(8)

where α and β are weight factors of the signaling cost and the packet delivery cost to adjust weights of both costs (α + β = 1). Binding refresh signaling cost considered in [15] is omitted in this analysis, since FMIPv6 is only performed when the handover takes place. The symbols of system parameters used in this section is given in Table 1. 3.3.1 The signaling cost The signaling cost is summed by costs of transferred signaling messages during the FMIPv6 , consists of the predictive mode handover. The signaling cost of FMIPv6, Csignal Table 1 Symbols of system parameters Symbols

Meanings

α

Weight factor of signal cost

β

Weight factor of packet deliver cost



Weight factor of buffering in an AR

γ

Weight factor of retransmission

lc

Control packet size

ld

Data packet size

wl

Weight factor of wireless link

wd

Weight factor of wired link

hPN

Average hops between PAR and NAR

hAH

Average hops between AR and HA

λp

Packet arrival rate

tL2

The layer 2 handover delay

twl

Wireless transmission delay

twd

Wired transmission delay

666

S. Ryu et al. predict.

react. . cost, Csignal , and the reactive mode cost, Csignal predict.

FMIPv6 react. Csignal = (1 − Prf ) · Csignal + Prf · Csignal ,

(9)

where Prf is the PPMF. In FMIPv6, signaling costs of predictive mode and reactive mode can be calculated, referring to Fig. 1. The signaling costs in the predictive mode and the reactive mode consists of costs induced between layer 2 trigger and link down, and by IP connectivity process in the new network. Then signaling costs of the predictive mode and the reactive mode are the following:  predict. predict. predict.  , Csignal = μh · lc · Cprev. + Cnew   react. react. react. , = μh · lc · Cprev. + Cnew Csignal predict.

where Cprev.

(10) (11)

react. are costs induced between layer 2 trigger and link down in and Cprev. predict.

react. are costs the predictive mode and the reactive mode, respectively. Cnew and Cnew induced by IP connectivity process at the new network in the predictive mode and the reactive mode, respectively. predict.

= 4 · ωl + 3 · ωd · hPN ,

(12)

predict.

= ωl + 2(ωl + ωd · hAH ),

(13)

Cprev. Cnew

react. Cprev. = 3 · ωl ,

(14)

react. = 2 · ωl + 2 · ωd · hPN + 2(ωl + ωd · hAH ), Cnew

(15)

where lc is the average length of a control packet, ωl and ωd are weight factors of wireless and wired link, respectively. hPN and hAH are the average number of hops between a PAR and an NAR, and between an AR and a HA, respectively. 3.3.2 The packet delivery cost FMIPv6 , consists of delivery costs for the preThe packet delivery cost of FMIPv6, Cdelivery predict.

react. . dictive mode, Cdelivery , and the reactive mode, Cdelivery predict.

FMIPv6 react. = (1 − Prf ) · Cdelivery + Prf · Cdelivery . Cdelivery

(16)

In terms of the packet delivery cost, we consider the costs associated with packet forwarding and packet buffering. In FMIPv6, packets from correspondent nodes are forwarded to the MN through tunneling between the PAR and the NAR. Before the MN connects to the NAR, the forwarded packets are buffered at the PAR or the NAR. In the predictive mode, forwarding and buffering of packets are started after Tfast . In the reactive mode, if the layer 2 trigger is not occurred (Ttrigger ≤ 0), the packets are predict. react. are obtained as follows: lost during the handover. Then Cdelivery and Cdelivery predict.

packet

AR Cdelivery = μh · λp · Cforward · (THL − Tfast ) +  · μh · Cbuffer · (THL − Tfast ), (17)

Optimized fast handover scheme in Mobile IPv6 networks to support

 react. Cdelivery

=

packet

AR · T μh · λp · Cforward · THL +  · μh · λp · Cbuffer HL packet

μh · γ · Cforward · THL

667

(Ttrigger > 0) (Ttrigger ≤ 0), (18)

where λp is the packet arrival rate, ld is the average length of a data packet, and  packet and γ are weight factors of buffering in an AR and packet loss, respectively. Cforward which is a cost of forwarding a packet and is obtained as packet

Cforward = ld · (wd · hAH + wd · hPN + wl )

(19)

AR which is a cost of buffering in an AR obtained as Cbuffer AR = λ p · ld . Cbuffer

(20)

THL is the handover latency, Ttrigger is a delay between the layer 2 trigger event and the link down epoch. The handover latency is divided into three components: Ttrigger , tL2 which is the layer 2 handover latency, and Tnew which is a delay between the link up epoch and completion of registration. THL = Ttrigger + tL2 + Tnew .

(21)

Referring to Fig. 3, we calculate the handover latencies of two operation modes which predict. react. for the reactive mode: are THL for the predictive mode and THL predict.

THL

= Ttrigger + tL2 + 2 · twl + 2(twl + twd · hAH ),

react. = Ttrigger + tL2 + 2 · twl + 2 · twd · hPN + 2(twl + twd · hAH ). THL

(22) (23)

Also, wireless transmission delay, twl , and wired transmission delay, twd , are calculated referring to [15]:

l 1+q (24) twl (l) = + Lwl 1 − q BW l twd (l) =

l + Lwd + ω¯ q BW d

(25)

where l is a packet length, q is the probability of wireless link failure, ω¯ q is the average queueing delay at each router in the Internet [22], BW l and BW d are the bandwidth of wireless and wired links, respectively. Lwl and Lwd are wireless and wired link delays, respectively. 4 Analytical results In this section, we provide results from the performance evaluation of FMIPv6 based on the previous modeling to show its effectiveness for mobile cloud computing. Default values for parameters used in the performance evaluation are given in Table 2 [14, 15, 23, 24].

668

S. Ryu et al.

Fig. 3 FMIPv6 message flows and handover latency timelines

Table 2 Values of system parameters Symbols

Values

Symbols

Values

α

0.7

β

0.3



0.001

γ

2

lc

100 bytes

ld

1500 bytes

wl

10

wd

2

hPN

1

hAH

7

λp

10 packets/s

K

5

q

0.5

ω¯ q

0.5

Lwl

2 ms

Lwd

0.5 ms

BW l

11 Mbps

BW d

100 Mbps

tL2

50 ms

The previous studies have not considered network parameters to obtain values of the PPMF. They just regarded the value of the PPMF as a fixed one. Figure 4

Optimized fast handover scheme in Mobile IPv6 networks to support

669

Fig. 4 The signaling cost and the packet delivery cost depending on the PPMF

shows the effect of the PPMF on the signaling cost and the packet delivery cost. The signaling cost of the predictive mode is larger than that of the reactive mode, although the packet delivery cost of the predictive mode is smaller than that of the reactive mode. The signaling cost and the packet delivery cost of both the predictive mode and the reactive mode are not affected by the PPMF. The signaling cost of the combined mode decreases linearly as the PPMF increases, although the packet delivery cost of the combined mode increases linearly. This is because values of the PPMF increase linearly. When the PPMF is 0, the signaling cost and the packet delivery cost of the combined mode become those of the predictive mode, respectively. The signaling cost and the packet delivery cost of the combined mode become those of the reactive mode, respectively, when the PPMF is 1. This is because the combined mode consists of the predictive mode and the reactive mode through the PPMF. Then, the previous studies have used these values or arbitrary fixed values for the PPMF. However, the PPMF should be analyzed as the function of the radius of a cell, velocity of MNs and Tfast . In this paper, the PPMF is analyzed depending on those parameters. Figure 5 illustrates the PPMF as functions of the radius of a cell and Tfast . In Fig. 5(a), the PPMF decreases exponentially with the radius of a cell, while it increases proportionally with velocity of MNs. When the radius of a cell is small, the PPMF becomes large. This is because the overlapped area is affected by the radius of both cells. If the overlapped area is small, it is hard to transfer all of the additional signaling messages before the layer 2 handover is completed. That is, the PPMF becomes larger in small overlapped area than in large overlapped area. In addition, the PPMF becomes almost 1, when the radius of a cell approaches to 0. In Fig. 5(b), the PPMF increases as Tfast increases. Especially, the PPMF is heavily affected by Tfast , when the radius of a cell is small. As illustrated in Figs. 5(a) and 5(b), the radius of a cell and Tfast are considered as important parameters to obtain the PPMF. Thus, in order to evaluate performance of FMIPv6 in real network environments, various values for the radius of a cell and Tfast should be used to calculate the signaling cost and the packet delivery cost. The combined mode as well as the predictive mode and the reactive mode are analyzed in order to demonstrate the effect of the PPMF. For the evaluation of the

670

S. Ryu et al.

Fig. 5 The PPMF depending on radius of a cell and Tfast

signaling cost and the packet delivery cost, we assume that Tfast is the same as Ttrigger to demonstrate the effect of varying Tfast . If Ttrigger < Tfast , the PPMF cannot be obtained by network parameters, since the PPMF is always 1 in this case, referring to (4). The signaling cost is depicted in Fig. 6 as a function of Tfast . Figures 6(a)–6(d) show the results when the radius of a cell varies 10 m, 40 m, 70 m, and 100 m, respectively. The signaling cost of the reactive mode is smaller than that of the predictive mode, since the number of signaling messages during the handover in the predictive mode is higher than that in the reactive mode. We observe that the signaling cost of the combined mode decreases exponentially with Tfast , while costs of the predictive mode and the reactive mode are not affected by Tfast . This is because Tfast affects the PPMF. When Tfast increases, the signaling cost of the combined mode approaches that of the reactive mode. In the case that the radius of a cell is small, the signaling cost of the combined mode is heavily affected by Tfast , since the PPMF is heavily affected by Tfast . Thus, in order to reduce the signaling cost of the combined mode, Tfast and the radius of a cell should become large. Figure 7 demonstrates the handover latencies depending on the PPMF and Tfast , respectively. Generally, the handover latency of the predictive mode is smaller than that of the reactive mode. In Fig. 7(a), the predictive mode and the reactive mode are not affected by the PPMF, although the handover latency of the combined mode increases linearly as the PPMF increases. The handover latency of the combined mode becomes that of the predictive mode when the PPMF is 0, although it becomes that of the reactive mode when the PPMF is 1. This is because the combined mode consists of the predictive mode and the reactive mode through the PPMF. In Fig. 7(b), the handover latencies of each mode increase linearly as Tfast increases. The handover latencies of each mode are equally affected by Tfast , since latency values which are obtained by (22) and (23) are almost same. Figure 8 represents the effect of Tfast on the packet delivery cost. The packet delivery cost is largely affected by the handover latency. Generally, the packet delivery

Optimized fast handover scheme in Mobile IPv6 networks to support

671

Fig. 6 The signaling cost depending on Tfast

cost of the reactive mode is larger than that of the predictive mode, since the handover latency of the reactive mode is longer than that of the predictive mode. In Fig. 8, the packet delivery cost of the reactive mode is proportional to Tfast , while that of the combined mode is exponential with Tfast . However, the packet delivery cost of the predictive mode remains constant, since tunneling and buffering of packets are started after Tfast in the predictive mode. That is, in the predictive mode, Tfast is not included in the handover latency shown in (17). When Tfast is small, the packet delivery costs become small. The packet delivery cost of the combined mode approaches that of the predictive mode, as Tfast decreases. Figures 8(a)–(8d) show the results when the radius of a cell varies 10 m, 40 m, 70 m, and 100 m, respectively. The packet delivery cost of the combined mode is greatly affected by Tfast when the radius of a cell is small. When the radius of a cell is large, the packet delivery cost of the combined mode increases gradually. Thus, in order to reduce the packet delivery cost of the combined mode, Tfast should become as small as possible, while the radius of a cell should become large. On the other hand, to reduce the signaling cost, Tfast should become large. However, Tfast should become small to reduce the total cost,

672

Fig. 7 The handover latency depending on the PPMF and Tfast

Fig. 8 The packet delivery cost depending on Tfast

S. Ryu et al.

Optimized fast handover scheme in Mobile IPv6 networks to support

673

Fig. 9 The total cost as a function of layer 2 trigger time

since the total cost composed of the signaling cost and the packet delivery cost is heavily affected by the packet delivery cost. As mentioned in the above section, FMIPv6 is initiated by the layer 2 trigger. The layer 2 trigger informs the layer 3 of the layer 2 event. In the handover process, the layer 2 event can occur due to a change in the signal strength of one or more access networks [14]. The exact time when the layer 3 is notified of the layer 2 event should be determined to minimize the total cost. Thus, Ttrigger is considered with the effect of the changing Tfast to demonstrate the optimized total cost of the combined mode, while in the previous analytical results, Ttrigger is the same as Tfast . Figure 9 shows that the total costs depend on Ttrigger . The total costs increase, as Ttrigger increases. The total cost of the reactive mode is larger than that of the predictive mode, since the total cost is heavily affected by the packet delivery cost. When Ttrigger < Tfast , the total cost of the predictive mode is zero, while it jumps when Ttrigger ≥ Tfast . This is because the PPMF becomes 1 when Ttrigger < Tfast . Then, the total cost of the combined mode becomes the same as that of the reactive mode, when Ttrigger < Tfast .

674

S. Ryu et al.

In Figs. 9(a)–9(d) shows the results when Tfast varies 50 ms, 100 ms, 200 ms, and 400 ms, respectively. When Tfast is small, the total cost of the combined mode approaches that of the predictive mode, since the PPMF becomes small. Generally, the total cost of the combined mode becomes minimum when Ttrigger = Tfast . However, in the case that Tfast is longer than approximately 60 ms in Fig. 9, the total cost of the combined mode becomes minimum when Ttrigger is almost zero.

5 Conclusion In cloud computing, mobile networks should be efficiently managed to support seamless mobility to mobile users who request real-time services. To manage efficient mobile networks, FMIPv6 which is an extension of MIPv6 must be utilized. Several studies have been done to analyze the performance of FMIPv6 in different network environments. However, the previous works neither consider the PPMF, nor use various network parameters for the PPMF. In addition, in the most previous papers, two operation modes of FMIPv6 have not been analyzed altogether. In this paper, FMIPv6 combining two operation modes is analyzed considering the PPMF which is affected by the radius of a cell, Tfast which is the time required to transferring additional layer 2 signaling, and Ttrigger which is the layer 2 triggering time. In performance analysis, the total cost composed of the signaling cost and the packet delivery cost is investigated. Generally, when the radius of a cell is large, the signaling cost and the packet delivery cost of the combined mode becomes small. When Tfast is long, the signaling cost of the combined mode becomes small. However, the packet delivery cost of the combined mode becomes large when Tfast is long. Tfast should become as short as possible, since the total cost is heavily affected by the packet delivery cost. Then, Tfast and Ttrigger should be considered to reduce the total cost of the combined mode. Tfast is a predictable value, since it can be obtained by network parameters such as wireless and wired transmission delays. Then Ttrigger should become Tfast when Tfast is long while it should become almost zero when Tfast is short. Finally, FMIPv6 is optimized by using network parameters in this paper. The optimized FMIPv6 can be used in mobile networks for cloud computing, since FMIPv6 is a representative protocol of extensions for MIPv6. Then cloud computing will be able to provide seamless mobility services to mobile users efficiently.

References 1. Wei G, Vailakos A-V, Zheng Y, Xiong N (2009) A game-theoretic method of fair resource allocation for cloud computing services. J Supercomput. doi:10.1007/s11227-009-0318-1 2. Dikaiakos M-D, Pallis G, Katsaros D, Mehra P, Vakali A (2009) Cloud computing: distributed Internet computing for IT and scientific research. IEEE Internet Comput 13(5):10–13 3. Abu-Tair M, Min G, Ni Q, Liu H (2009) An adaptive medium access control scheme for mobile ad hoc networks under self-similar traffic. J Supercomput. doi:10.1007/s11227-009-0324-3 4. Li C, Li L (2009) Tradeoffs between energy consumption and QoS in mobile grid. J Supercomput. doi:10.1007/s11227-009-0330-5 5. Cherry S (2009) Forecast for cloud computing: up, up, and away. IEEE Spectrum 46(10):68

Optimized fast handover scheme in Mobile IPv6 networks to support 6. 7. 8. 9. 10.

11.

12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

675

Kruys J (2009) Untethered clouds. IEEE Wirel Commun 16(4):8–10 Golmie N (2009) Seamless mobility: are we there yet? IEEE Wirel Commun 16(4):12–13 Johnson D, Perkins C, Arkko J (2004) Mobility support in IPv6. IETF, RFC 3775 Koodli R (ed) (2005) Fast handovers for Mobile IPv6. IETF RFC 4068 Perez-Costa X, Torrent-Moreno M, Hartenstein H (2003) A performance comparison of Mobile IPv6, hierarchical Mobile IPv6, fast handovers for Mobile IPv6 and their combination. ACM Mobile Comput Commun Rev 7(4):5–19 Gwon Y, Kempf J, Yegin A (2004) Scalability and robustness analysis of Mobile IPv6, fast Mobile IPv6, hierarchical Mobile IPv6, and hybrid IPv6 mobility protocols using a large-scale simulation. In: Proc IEEE int conf on commun (ICC’04), vol 7, pp 4087–4091 Li R, Li J, Wu K, Xiao Y, Xie J (2008) An enhanced fast handover with low latency for Mobile IPv6. IEEE Trans Wirel Commun 7(1):334–342 Fathi H, Chakraborty S, Prasad R (2007) Optimization of Mobile IPv6-based handovers to support VoIP services in wireless heterogeneous networks. IEEE Trans Veh Technol 56(1):260–270 Pack S, Choi Y (2003) Performance analysis of fast handover in Mobile IPv6 networks. In: Proc LNCS, vol 2775, pp 679–691 Makaya C, Pierre S (2008) An analytical framework for performance evaluation of IPv6-based mobility management protocols. IEEE Trans Wirel Commun 7(3):972–983 Ho J-S-M, Akyildiz I-F (1995) Mobile user location update and paging under delay constraints. ACMBaltzer J Wirel Netw 1:413–425 Akyildiz I-F, Wang W (2002) A dynamic location management scheme for next-generation multitier PCS systems. IEEE Trans Wirel Commun 1(1):178–189 Markoulidakis J, Lyberopoulos G, Anagnostou M (1998) Traffic model for third generation cellular mobile telecommunications systems. ACM-Baltzer J Wirel Netw 4:389–400 Wan G, Lin E (1999) Cost reduction in location management using semi-realtime movement information. ACM-Baltzer J Wirel Netw 5:245–256 Lin Y-B (1997) Modeling techniques for large-scale PCS networks. IEEE Commun Mag 35(2):102– 107 McNair J, Akyildiz I-F, Bender MD (2000) An inter-system handoff technique for the IMT-2000 system. In: Proc IEEE INFOCOM, pp 208–216 McNair J, Akyildiz I-F, Bender MD (2001) Handoffs for real-time traffic in Mobile IP version 6 networks. In: Proc IEEE GLOBECOM, vol 6, pp 3463–3467 Xie J, Akyildiz I-F (2002) A novel distributed dynamic location management scheme for minimizing signaling costs in Mobile IP. IEEE Trans Mobile Comput 1(3):163–175 Lai WK, Chiu J.C (2005) Improving handoff performance in wireless overlay networks by switching between two-layer IPv6 and one-layer IPv6 addressing. IEEE J Sel Areas Commun 23(11):2129– 2137

Optimized fast handover scheme in Mobile IPv6 ... - Springer Link

Jun 12, 2010 - Abstract In the future cloud computing, users will heavily use mobile devices. Mo- bile networks for cloud computing should be managed ...

1MB Sizes 3 Downloads 296 Views

Recommend Documents

Macro/micro-mobility fast handover in hierarchical mobile IPv6
Abstract. Mobile Internet Protocol version 6 (MIPv6) has been proposed to solve the problem of mobility in the new era of Internet by handling routing of IPv6 packets to mobile nodes that have moved away from their home network. Users will move frequ

Multicast based fast handoff in Hierarchical Mobile IPv6 ...
Handoff-Aware Wireless Access Internet Infrastructure. (HAWAII) [15]. ... home agent by sending another BU that specifies the binding between its home address ...

Globally Optimized Spherical Point Arrangements - Springer Link
Pintér Consulting Services, Inc. (PCS) & Dalhousie University,. PCS: 129 ... of local optima to be some exponential function of the number of points: see, e.g., the.

Handover method for mobile radio system
Jan 11, 1999 - Nakajirna, A., Advanced Mobile Communication Network. 5,452,473 A .... is, inter alia, to enable the degree of coverage to be made greater Without the ...... ters BM and Bnb Which has the best radio transmission conditions ...

MULTICAST, HIERARCHICAL AND FAST HANDOVER ...
*Department of Computer and Communication System Engineering, Faculty of Engineering. **Department ... Laptop is used as Mobile Host (MH), two personal.

Handover method for mobile radio system
Jan 11, 1999 - IEEE Transaction on Vehicular Technology, vol. VT—19, No. 4,955,082 A ... Nakajirna, A., Advanced Mobile Communication Network. 5,452,473 A. 9/1995 ... Wireless Communications Research Institute, Ulm (Ger many), pp.

Fast Support Vector Data Description Using K-Means ... - Springer Link
Using K-Means Clustering. Pyo Jae Kim, Hyung Jin Chang, Dong Sung Song, and Jin Young Choi. School of Electrical Engineering and Computer Science,.

LNCS 4233 - Fast Learning for Statistical Face Detection - Springer Link
Department of Computer Science and Engineering, Shanghai Jiao Tong University,. 1954 Hua Shan Road, Shanghai ... SNoW (sparse network of winnows) face detection system by Yang et al. [20] is a sparse network of linear ..... International Journal of C

Handover based Backlogged Packet Discard Scheme ...
1. the statistics of the call arrival process ... A BS (base station) is assumed to be located at the center of ... From busy cycle analysis for an M/M/1/1 queue.

Experimental Study on Mobile RFID Performance - Springer Link
a realistic mobile setting. ... understanding about the reading performance for mobile scenario. ..... ysis of energy consumption for ISO 18000-7 RFID networks.

Topological Synthesis of Mobile Backbone Networks ... - Springer Link
topological synthesis and management of Mobile Backbone Network (MBN) based architectures. ... and video, notwithstanding the lack of a fixed infrastructure.

Topological Synthesis of Mobile Backbone Networks ... - Springer Link
topological synthesis and management of Mobile Backbone Network (MBN) .... accessibility (i.e., a certain percentage of nodes should be within a fixed small.

Optimized Mobile Search Engine - IJRIT
IJRIT International Journal of Research in Information Technology, Volume 1, .... So Many existing personalized web search systems are based click through data to .... And this design allows user privacy to be preserved in certain degree. Two.

Ambiguity in electoral competition - Springer Link
Mar 1, 2006 - How to model ambiguity in electoral competition is a challenge for formal political science. On one hand ... within democratic political institutions.1 The ambiguity of political discourse is certainly ...... Princeton University Press,

Exploring Cultural Differences in Pictogram ... - Springer Link
management applications such as Flickr and YouTube have come into wide use, allowing users to ... the meaning associated with the object. Pictorial symbols ...

Complexified Gravity in Noncommutative Spaces - Springer Link
Complexified Gravity in Noncommutative Spaces. Ali H. Chamseddine. Center for Advanced Mathematical Sciences (CAMS) and Physics Department, American University of Beirut,. Lebanon. Received: 1 June 2000 / Accepted: 27 November 2000. Abstract: The pre

Directional dependence in multivariate distributions - Springer Link
Mar 16, 2011 - in multivariate distributions, and introduce some coefficients to measure that depen- dence. ... and C(u) = uk whenever all coordinates of u are 1 except maybe uk; and. (ii) for every a = (a1, a2,..., ...... IMS Lecture Notes-Mono-.

Molecular diagnostics in tuberculosis - Springer Link
Nov 10, 2005 - species, detection of drug resistance, and typing for epi- demiological investigation. In the laboratory diagnosis of tuberculosis, the nucleic acid ...

Ethics in agricultural research - Springer Link
improvement criteria (Madden 1986). Some see distributional problems as reason to reject utilitarianism entirely (Machan 1984; Dworkin I977). Each of these.

Visceral regeneration in the crinoid - Springer Link
sic characteristic of life, although it can be lost when their costs are higher than their ... individuals with visceral regeneration in progress [7, 25–28], indicates that the ... In the following stages, the regrowth of the intestinal tract can i

Management of Diabetes in Pregnancy - Springer Link
Dec 3, 2011 - profound effects on multiple maternal organ systems. In the fetus, morbidities ... mellitus . Metformin . Glyburide . Pregnancy; diabetes management. Clinical Trial Acronyms. ACHOIS Australian Carbohydrate Intolerance Study in. Pregnant

This Month in APR - Springer Link
projected by measuring the node degree, betweenness and closeness for each node within these networks. Additionally, promiscuity maps and heat maps were.

Path delays in communication networks - Springer Link
represent stations with storage capabilities, while the edges of the graph represent com- ... message time-delays along a path in a communication network.