1

A Simple Remedy for Idle Mode via Proxy MIP Sunggeun Jin, Student Member, IEEE, Chulsik Yoon, Member, IEEE, and Sunghyun Choi, Senior Member, IEEE

Abstract—While recent wireless networks begin to support Mobile Internet Protocol (MIP), even the latest wireless network technologies such as IEEE 802.16 Wireless Metropolitan Area Networks (WMANs), IEEE 802.11 Wireless Local Area Networks (WLANs), and High-Speed Downlink Packet Access (HSDPA) are limited in exploiting idle mode under mobile environments in which MIP is employed since they are designed without considering MIP combined with idle mode. In other words, an idle Mobile Station (MS) should perform MIP handoffs despite that it cannot receive anything while staying in idle mode, and hence, it incurs inexpedient operations. We propose a simple protocol, which makes idle mode integrated with MIP seamlessly by utilizing Proxy MIP (PMIP). It processes MIP handoffs on behalf of idle MS, thus allowing idle MS to stay in idle mode consistently.

I. I NTRODUCTION For a wide acceptance of Voice over IP (VoIP) service in recently emerging wireless networks such as 802.16e WMANs, 802.11 WLANs, and HSDPA systems for highspeed wireless Internet access, the standby time extension of battery-powered VoIP phones is very critical. Idle mode in the wireless networks was proposed to enhance power saving efficiency, thus ultimately leading to longer standby time as well as to provide a key benefit to reduce signaling cost1 while MS remains without any active session. IP mobility management is also regarded indispensable to VoIP service in a mobile environment. However, wireless networks deal with only Medium Access Control (MAC) and PHYsical (PHY) layers so that it is not clearly specified how to manage IP mobility while MS is in idle mode. For this reason, in [1], [2], the authors began to discuss this issue by proposing IP paging schemes to integrate IP layer with idle mode. However, as detailed later, the proposals regarding IP paging, which have been developed so far, are not optimized enough to achieve the benefits which idle mode targets at. The reason comes from the fact that idle MS should wake up either to conduct MIP handoff or to support IP paging for which it cannot help receiving broadcast/multicast IP packets since both rely on IP broadcast/multicast. Consequently, idle MS is forced to require more signaling cost and more power consumption. Sunggeun Jin and Chulsik Yoon are with ETRI, Daejeon 305-700, Korea. Sunghyun Choi is with School of Electrical Engineering and INMC, Seoul National University, Seoul 151-744, Korea. This work was in part supported by the MKE (Ministry of Knowledge Economy), Korea, under the ITRC program (IITA-2008-C1090-0801-0013) and IT R&D program (2007-F-038-02, Fundamental Technologies for the Future Internet) supervised by IITA. 1 In this letter, signaling cost is defined as the amount of resources required to manage mobility.

Fast emerging Proxy MIP (PMIP) technology brings a new opportunity for idle mode. PMIP is originally designed to conduct IP handoff on behalf of MS incapable of doing so. In fact, it is more suitable for idle mode since neither idle MS has an active VoIP session nor it needs to perform IP handoff for a proper VoIP-packet routing. Though the ongoing PMIP standard activity is one of the hottest in Internet Engineering Task Force (IETF) [3], PMIP does not represent a specific standard technology, but commonly implies an IP layer component to perform IP layer handoff on behalf of MS. In this letter, we propose a simple generic protocol for idle mode by employing PMIP. When an MS enters idle mode, PMIP assumes its role of conducting IP handoff procedure until the end of idle mode via a successful paging. The proposed protocol is so simple that it can be easily employed by currently deployed wireless networks. II. R ELATED W ORK AND P ROPOSAL A. Related Work According to [1], wireless network pages idle MS when an agent capable of triggering paging procedure receives a newly-arriving IP packet destined to the idle MS. The proposal was the first trial to integrate heterogeneous wireless networks. However, the authors do not address how to cope with the case that idle MSs move across the boundaries, where MIP handoffs are necessary. Subsequently, a technical gap is left in the case for which they should have provided a remedy. In [2], the authors first proposed IP paging scheme, which broadcasts/multicasts IP packets to page idle MS. Nevertheless, their proposal has a limitation. Since wireless networks generally do not provide any policy for idle MS to differentiate the packets for IP layer paging and those for MIP handoff among broadcast/multicast IP packets over the air, the idle MS has to receive all broadcast/multicast IP packets. It implies that their IP paging scheme is not optimized for idle MS so that it has to wake up to receive broadcast/multicast IP packets, thus incurring redundant power consumption as well as signaling cost. In [4], [5], the authors propose a complementary paging scheme without IP paging for the 802.11 and the 802.16e, respectively. They suggest that MIP handoff should be postponed until the completion of idle mode and the authors of [4] refer to their proposed MIP handoff as delayed handoff. Under the delayed handoff, an idle MS can save power by not receiving broadcast/multicast IP packets. Moreover, it generates less signaling cost because it performs MIP handoff only once. On the other hand, since routing information remains unchanged due to the delayed handoff, new IP packets destined to the

2

paged MS arrives at the old Foreign Agent (FA), where an idle MS entered idle mode, and then, they travel to the new FA under which the idle MS performed delayed handoff. Therefore, their proposals cause long startup latency,2 which is generally proportional to the distance that an idle MS has moved during its corresponding idle time. Along with the proposal, the authors of [5] provide another approach that an idle MS performs MIP handoff whenever it changes its FA, and hence, it should be paged to wake up to receive broadcast/multicast IP packets in order to conduct MIP handoff. In consequence, it is inefficient in both signaling cost and power consumption, but the startup latency can be minimized. In this letter, we refer to this proposal as immediate handoff. As expected, long distance between the FAs, where MIP handoffs are conducted, helps saving signaling cost at the cost of long startup latency. For this reason, there exists a tradeoff relationship between the signaling cost and the startup latency as a function of the distance. B. Proposed Scheme We design FA to perform MIP handoff on behalf of an idle MS, which is referred to as PMIP procedure when an idle MS enters FA’s coverage. Idle MS neither performs MIP handoff nor receives broadcast/multicast IP packets, thus reducing power consumption as well as signaling cost significantly. In fact, FAs, however, do not need to conduct PMIP procedure whenever idle MS enters their coverage since idle MS does not have a session. We assume FAs perform PMIP procedure only if an idle MS reaches the coverage of a new FA located at a predefined distance from an old FA (oFA), which had performed the latest PMIP procedure. For better understanding, we introduce a protocol based on the 802.16e wireless network as follows: (1) An MS enters idle mode by completing idle mode entrance procedure involved with a Paging Controller (PC). The PC begins managing the location of the MS after activating PMIP in the FA connected with the Base Station (BS) under which the MS enters idle mode. (2) When the idle MS moves to the coverage of a new FA, it should conduct location update procedure with the PC through a BS according to the 802.16e standard. In this case, the PC3 obtains the distance between new FA and an oFA. If the distance is equal to a predefined value, it requests the new FA (nFA)4 to conduct PMIP procedure. Additionally, it commands the oFA to remove obsolete context about the idle MS. (3) Since FAs do not perform PMIP procedure whenever idle MS enters new FA-coverage, IP packets destined for the idle MS arrive at oFA. For this reason, the oFA triggers the PC to page the idle MS. Consequently, the paged idle MS can conduct MIP handoff by itself after the completion of a network entry procedure for the 2 Startup latency represents a duration from an instance that a Paging Controller (PC) begins to page an idle MS until the paged MS receives the first IP packet. 3 We can design PC to have a mapping table containing distances between FAs based on pre-configured location information for each FA. 4 nFA represents the new FA performing PMIP procedure.

wireless network while oFA forwards the arrived IP packets to the FA under which the paged MS is. As a result, our proposal enables idle MS to fully exploit the power saving efficiency, which idle mode can provide by removing the necessity to receive broadcast/multicast IP packets. III. A NALYSIS AND R ESULTS A. Cost Function As discussed earlier, the tradeoff relationship between the signaling cost and the startup latency is a function of a distance (d) between nFA and oFA. It can be represented by a metric, (CF (d)), called cost function, including both the signaling cost (CH (d)) and the startup latency (CL (d)). CF (d) = aCH (d) + bCL (d),

(1)

where a and b are weighting factors to conform the unit for each term, and hence, CF (d) becomes a constant without unit. We refer to a distance minimizing CF (d) as optimal distance (d∗ ). In this equation, the cost for the delayed handoff is denoted by an extreme case of our protocol with d set to the number of FA boundary crossings for its entire idle time. Immediate handoff can approximate to another extreme by letting d = 1 without considering the clear difference regarding the overhead required for an idle MS to perform MIP handoff by itself. B. Numerical Analysis In our analysis, we assume bidirectional random walk model, which is powerful enough to represent the tradeoff relationship, due to its simplicity. Under this model, an MS moves on a path in a back-and-forth manner with equal probabilities. To our best knowledge, there has never been a correct equation for distance-based mobility management strategies, which our proposal is classified into. For example, the authors of [6] missed that MS does not conduct handoff in case that it returns to the place where it did once before it moves beyond an optimal distance. (n) Let P(d,k) be the probability that PMIP procedure is conducted more than n times under the condition that an MS can cross FA boundaries k times with a given value of d for its entire idle time. We simply call this condition the movement (n) condition. For d ≥ 1, k ≥ 0, and n ≥ 0, we derive P(d,k) by: (n)

P(d,k)

Pb(k−d)/2c (n−1)  ( 12 )2i+d Q(d,i) P(d,k−(2i+d)) , d < k,  i=0 (n−1) d 1 = 2 × ( 2 ) P(d,k−d) , d = k,   0, d > k,

(2)

Q(d,i) = (3)  2, i = 0,   ³¡ ¢ ¡ ¢´  ¡2i¢ Pb(d−1)/2c (d) 2i 2i 2 i + l=1 w(d−2l) i+l + i−l , i ≤ bd/2c, ´ ³  P  (2i) (2i) 2q (2i) + b(d−1)/2c w(d) l=1 (d−2l) q(2i,l) + q(2i,−l) , i > bd/2c, (2i,0) (0) (m) (m−1) (m−1) (m) where P(d,k) = 1, w(r) = w(r+1) + w(r−1) , and w(0) = 0. It (m) proves true that w(r) = m − 2 only if r = m − 2. Additionally, ¡x¢ (j) q(x,y) = y for j = d, and otherwise, for j > d, it can be obtained recursively by:

3

the movement condition. Under the same condition, we obtain (n) simulation results denoted by s(d,k) , and validate our analy¯ ¡ ©¯ (n) ¯ª¢ < 0.004, sis by confirming that max ∀e ∈ ¯p(n) (d,k) − s(d,k) where 1 ≤ d < 100, 1 ≤ k < 100, and 0 ≤ n < 100. (n) Set Ψ(d,k) consists of the elements representing the number of FA boundaries which an MS can cross until a new call arrival after PMIP procedures are performed only n times 0 under the movement condition. Its element ψ(i, ψ )(> 0) is ¡ ¢ 0 0 indexed by both i and ψ , where 0 ≤ i ≤ max 0, b ψ 2−d c and ψ 0 > 0. Recursively, ψ 0 is an element of a set Ψ(n−1) (d,k) . (n) Consequently, we derive Ψ by: (d,k) n o (n)

(n−1)

Ψ(d,k) = ψ(i, ψ 0 )|ψ(i, ψ 0 ) = ψ 0 − (2i + d), ψ 0 ∈ Ψ(d,k)

, (5)

(0) Ψ(d,k)

where the initial condition = {k}. (n) c(d,k) represents the number of FA boundaries between an oFA and the FA where an idle MS is paged under the movement condition, and hence, can be used to obtain the startup latency. We derive c(n) (d,k) by: (n)

c(d,k) =

X

µ ¶k ψ X 1 (n) L(d,ψ) u(i) · r(i, ψ), 2

(6)

i=−ψ (n−1) Q(d,ψ(i,ψ0 )) L(d,ψ0 )

(n) ∀ψ∈Ψ(d,k)

where L(n) (d,ψ(i,ψ 0 )) = equation, r(i,  ψ) is given by:

and L(0) (d,k) = 1. In this

 |i|, for ψ = even and i = even, (7) u(i) = |i|, for ψ = odd and i = odd,  0, otherwise,   |i| ≥ d, 0, (8) r(i, ψ) = 1, |i| = ψ,  r(ψ − 1, i − 1) + r(ψ − 1, i + 1), otherwise. Let CH (d, k) and CL (d, k) be the cost and the latency for CH (d) and CL (d), respectively, under the movement condition.

Accordingly, we derive the both as follows: CH (d, k) = Ch

k X

(n)

np(d,k) ,

(9)

n=0

CL (d, k) = Cl

k X

(n)

c(d,k) ,

(10)

n=0

where Ch and Cl are system-dependent constants regarding the PMIP procedure and the startup latency, respectively. CH (d, k) is a monotonically decreasing function while CL (d, k) is a monotonically increasing function since CH (d, k) ≤ 0 and CL (d, k) ≥ 0, where 2 < d < 100 and 2 < k < 100. If there exists an integer value of d∗k such that ∆CF (d, k) = 0, it is an optimum minimizing CF (d, k). Otherwise, d∗k =argmin (CF (d, k), CF (d + 1, k)) such d

that ∆CF (d, k) < 0 and ∆CF (d + 1, k) > 0. In order to obtain d∗k , we adopt a simple linear searching algorithm: let d = 3; while (∆CF (d, k) ≤ 0 and d 6= k) d = d + 1; d∗k = argmin (CF (1, k), CF (2, k), CF (d − 1, k), CF (d, k)); d

18 16 Analysis (r=0.1) Analysis (r=1) Analysis (r=10) Simulation (r=0.1) Simulation (r=1) Simulation (r=10)

14 Optimal distance

 (j−2) d (j−2) 1  + q(x,y−1) , y = b(d − 1)/2c,   2 (3 + (−1) )q(x,y)   1 (3 + (−1)d )q (j−2) + q (j−2) , y = −(b(d − 1)/2c), (j) (x,y) (x,y+1) q(x,y) = 2 (j−2) (j−2) (j−2)  y 6= ±(b(d − 1)/2c), 2q(x,y) + q(x,y+1) + q(x,y−1) ,   2q (j−2) , y = 0 and d = 2. (x,y) (4) (n) (n) (n+1) From Eqs. (1)–(4), we define p(d,k) =P(d,k) − P(d,k) , which is the probability that PMIP is performed exactly n times under

12 10 8 6 4 2

2

Fig. 1.

10

20

30

40 50 60 70 Number of FA boundary crossings (=k)

80

90

Optimal distance where FA should conduct PMIP procedure.

C. Results Fig. 1 shows optimal distances (d∗k ) when an idle MS crosses FA boundaries k times during its idle time. The values of d∗k oscillate since an MS reaches different FA depending on whether k is even or not in the bidirectional random walk model. By comparing analytic results with simulation results, we confirm that our analytical model is correct. ¡ ¢ h We introduce a ratio r = aC representing relative imbCl portance between the signaling cost and the startup latency. In case that r = 10, delayed handoff is partially effective only if k ≤ 11 since the latency becomes considerable for larger k. However, as r increases, delayed handoff becomes effective for a wider range of k. When r = 1, d∗k increases logarithmically, and is approximately blog2 (3k)c. Therefore, it can be determined easily. In case that r = 0.1, the startup latency dominates the signaling cost, and hence, d∗k should be small. For this reason, immediate handoff is more beneficial for r ¿ 0.1. IV. C ONCLUSION We propose a new protocol integrating PMIP into the existing wireless networks, which enables an idle MS to fully exploit the idle mode. Moreover, we provide an insight to utilize the proposed protocol depending on the tradeoff relationship between the signaling cost and the startup latency. R EFERENCES [1] R. Ramjee et al., “IP Paging Service for Mobile Hosts,” in Proc. ACM MOBICOM’01, July 2001. [2] X. Zhang, J. G. Castellanos, and A. T. Campbell, “P-MIP: Paging Extensions for Mobile IP,” Mobile Networks and Applications, April 2002. [3] J. Kempf, “Goals for Network-Based Localized Mobility Management (NETLMM),” IETF RFC 4831, Apr. 2007. [4] S. Jin, K. Han, and S. Choi, “A Novel Idle Mode Operation in IEEE 802.11 WLANs,” in Proc. IEEE ICC’06, June 2006. [5] J. Na et al., “Two Alternative Registration and Paging Schemes for Supporting Idle Mode in IEEE 802.16e Wireless MAN,” in Proc. IEEE VTC’06 Fall, Sep. 2006. [6] J. Cao et al., “Design and Performance Evaluation of an Improved Mobile IP Protocol,” in Proc. IEEE INFOCOM’04, Mar. 2004.