A Tiered Mobility Management Solution for Next Generation Wireless IP-Based Networks

Transcription

1 A Tiered Mobility Management Solution for Next Generation Wireless IP-Based Networks I. S. Misra, Member, IEEE, M. Chakraborty, D. Saha, Senior Member, IEEE and A. Mukherjee, Senior Member, IEEE I. S. Misra is with the Department of Electronics and Telecommunication Engineering, Jadavpur University, Calcutta , India. M. Chakraborty is with Netaji Subhash Engineerin College, Calcutta-152, India D. Saha is with the MIS and Computer Science Group, IIM Cacutta, India A. Mukherjee is with IBM India Private Limited BCS Salt Lake, Calcutta , India ABSTRACT This article describes an optimal mobility management solution for next generation wireless IP-based networks. It has been found earlier that hierarchical architectures prove advantageous in minimizing signaling overhead, but our ultimate goal is to find out the optimum hierarchy level, which should provide best performance in terms of networks parameters like signaling overhead, handoff latency and frequency of location updates. For this purpose an n- tiered architecture has been considered and optimality test on the number of tiers has been performed. Analytical and simulation (using ns-2) results have been provided to establish that three-tiered architecture is the optimal one. A comparative analysis of the proposed three-tiered architecture called Three Level Mobility Model (TLMM) with other existing hierarchical network architectures forms an important aspect of this paper. Key words: foreign agent, gateway mobility agent, handoff latency, location update, optimal hierarchy, mobility management, signaling overhead I. INTRODUCTION The present era of computing has ushered in rapid growth of wireless networks fuelled by widespread use of portable computers and by the popularity of the Internet. Seamless user mobility has become the need of the hour. Hence, networks of tomorrow have to be robust enough to deal efficiently different mobility patterns of the users. Many protocols have 8 China Communications December 2006

2 been designed and implemented that support IP mobility. Among them Mobile IPv4 (MIPv4) [1], MIPv4 Regional Registration (MIPv4 RR) [2], MIPv6 [3], [4] and Hierarchical MIPv6 [5] are noticeable. MIPv4 requires a Mobile Node (MN) to have a static Home Agent (HA) and a permanent home IP address. This is not always desirable for MNs, in commercial Internet Service Provider (ISP) domain. This is because here most ISPs provide automatically allocated temporary IP address to users via Dynamic Host Configuration Protocol (DHCP) or Point-to-Point Protocol (PPP). Furthermore, for some big network domains, the MN s current network attach point could be far away from the static HA and hence could cause severe triangular routing problems. Route optimization takes the edge off this problem but it still requires substantial remote HA signaling and does not help much when the MN is moving very fast. It also imposes high signaling overhead on the Correspondent Node (CN) for processing binding update messages and data encapsulation. Finally, for a fast moving MN, when it keeps migrating to a nearby Foreign Agent (FA), which is typical for wireless network subscribers, the signaling with the remote HA will cause an unacceptable long delay. MIPv4 RR uses a hierarchy of FAs in a visited domain so as to efficiently handle local movements of MNs without the need of frequent registrations with the HA [6]. The idea here is that the movement of an MN within the visited domain and in particular under one globally routable entity, denoted as the Gateway Foreign agent (GFA), is hidden from the HA so that the number of signaling messages to the home network, in conjunction with the time needed for the MN to update the path to its current location is reduced. This is because a change on the MN s path within a visited network is not propagated to the HA or the CN but is handled locally. This proposal uses a tree-like hierarchy of FAs where the FA located at the root of the tree is called the GFA. It is possible that several levels of regional FAs (RFA) are supported between the GFA and the lowest-level FAs. After an MN moves to new foreign domain, it registers with its home network where it uses the IP address of the GFA as its care-of-address (CoA). In MIPv6, the MN is obliged to send binding update massage to its CNs and HA each time it changes its point of attachment. This causes significant processing overhead as the number of MNs increases. In addition, handoff-speed performance is aggravated because the MN waits for an end-to-end path establishment so that it can receive packets on the new access router (AR). To overcome these limitations, Hierarchical MIPv6 (HMIPv6) introduced a local entity within the access network, the Mobility Anchor Point (MAP), which can be located at any level in a hierarchy of routers, including the AR. The idea is that the movement of the MN within MAP domain is not visible to the CNs and HA. So the latter need not be notified for MN movements within the MAP s subnet. The MAP intercepts the packets destined for MN and tunnels them to the actual location of the MN [7]. Review of the different hierarchical protocols show that increasing the number of hierarchical levels proves advantageous in the performance of signaling overhead but deficiently deteriorates the handoff latency and frequency of location updates as the MN moves very fast across multiple subnets in a domain. The performance further degrades with the increase in the number of MNs [8]. So in this paper we investigate for a hierarchical protocol architecture that should provide most favorable handoff latency, signaling overhead and frequency of location updates in an IP-based network scenario. For this purpose we consider an n-tiered architecture and execute optimality test on the number of tiers. The rest of the paper is divided as follows. After an introduction in section I, the analytical model of an n-tiered architecture is presented in section II. Section III provides the optimality test results. Both analytical as well as simulation results using ns-2 support the supremacy of the three-tiered architecture. Section IV provides an overview of the Three Level Mobility Model (TLMM), the proposed three-tiered architecture. This section also provides comparative performance analyses of the different hierarchical architectures. Finally section V concludes the paper with some highlights on future works. China Communications December

3 II. ANALYTICAL MODEL OF TIERED ARCHITECTURE For the purpose of finding out the optimal tiered mobility model in an IP-based wireless network let us define few terms with a view to fulfilling the performance criteria of any network system. The important network parameters of interest are location update frequency, handoff latency and signaling overhead. It is well known that terminal mobility management in wireless network environment consists of two components: location management and handoff management [9]. Location management is taken care of by location update (LU) or registration, which is a process that enables a network to discover the current point of attachment of an MN for information delivery. An MN is tracked through LU process in which it informs the network of its location at times triggered by movement, timer expiration, and so on. Locating an MN is performed through search procedures, when the network pages the MN. There is a trade-off between how closely the network tracks the current location of an MN, vs. the time and complexity required to locate an MN whose position is not precisely known. Handoff management enables a network to maintain a connection as an MN continues to move and change its point of attachment to the network. Signaling overhead is the total load on the network in the entire process of LU and handoff [10]. Accordingly we require having three different types of analytical models as required by the above three network parameters. We consider an n-tiered hierarchical mobility model, where there are several layers of FAs i.e., layer-1 FAs (L 1 FA) at the lowest layer, L 2 FAs at layer-2 and so on. The FAs at one layer are under the control of FAs immediately above their layer. So L 1 FAs are under the scope of L 2 FAs, L 2 FAs are under the scope of L 3 FAs and so on. So we define here a tree-like hierarchy of FAs, where FA located at the root of the tree (nth layer) is called the Root FA (RFA). A. Location update model The LU frequency (LU F ), which is the number of occurrences of registrations, may be expressed by Fig. n-tiered architectural overview 10 China Communications December 2006

4 the following equation: LU F = p * [f 1 /(x 1* x 2* x 3*... * x n-1 )] (1) where p: number of MNs f 1 : number of subnets (L 1 FAs) x 1, x 2, x 3,...,x n-1 : number of FAs under layer-2, layer-3,...layer-n FAs respectively. Here we make an assumption that x 1 >> x 2 >> x 3 >>...,x n-1 as the architecture follows a tree-like structure [11]. Whenever an MN attaches to a foreign domain in an n-tiered architecture, it has to acquire n- levels of CoAs starting from the lowest layer FA to the RFA as shown in Fig.1. It registers the RFA-CoA with its HA. After initial registration with HA, the mobility of the MN is handled locally so long it is under the same RFA. If, however the host changes its RFA, it has to register with HA. If, on the other hand the MN continues to move to different L 1 FAs under the same L 2 FA, it has to register with the latter. If the MN changes its L 2 FA, registration has to be made to L 3 FA and so on. So HA need not be notified of the current point of attachment of MN until eventually the MN changes its RFA. B. Signaling overhead model The expression for Per-Hop Signaling Overhead (SO PH ) for an n-tiered architecture is given below (Fig. 2). (2a) where l 1, l 2, l 3,...l n : length of the registration packets Fig. Registration/handoff process China Communications December

5 (in bytes) from RFA, L n-1 FAs, L n-2 FAs,..., L 1 FAs to HA respectively, which are defined as follows. l 2 = l 1 + (4 * 1), l 3 = l 1 + (4 * 2),..., l n = l 1 + (4 * (n-1)) So, l 1 <l 2 <...<l n Since l 1 involves only RFA CoA, its packet length is the least. l 2 involves two CoAs; that of RFA and l n- FA), l involves three CoAs (CoA of RFA, CoA of 1 3 l n-1 FA and CoA of l n -2FA) and so on. Each time an internal address is added the packet length increases by 4 bytes [12]. Here t 1... t n-2, t n-1, t n, average duration for which MN remains within the same L 1 FA, L 2 FA, L 3 FA,...RFA respectively are defined as follows. * * * * * t 2 = f 1 t 1, t 3 = f 2 f 1 t 1,..., t n = f n-1 f n-2... * f * * 2 f 1 t 1 The expression for Total Signaling Overhead (SO T ) including intermediate routers is given as follows. (2b) where h 1, h 2, h 3,...h n : number of hops ( routers) between L 1 FA and L 2 FA, L 2 FA and L 3 FA,...L n FA and HA respectively. C. Handoff model Regarding handoff latency we need to consider the timing aspects of signaling messages communicated. Global Handoff Latency (HL G ) for an n-tiered architecture may be expressed as follows (Fig. 2). HL G =T 1 +T T n (3a) Where, T 1, T 2,...T n : registration times from L 1 FA to L 2 FA, L 2 FA to L 3 FA...RFA to HA respectively for an n-tiered architecture. For onetiered, two-tiered, three-tiered architectures, t 1, t 2, t 3,... represent registration times from RFA to HA respectively. Local Handoff Latency (HL L ) for an n-tiered architecture up to k th level may be defined as follows. LH L = T 1 +T T k (3b) where T k is the registration time from L k-1 FA to L k FA. III. OPTIMALITY TEST RESULTS The main aim of this section is to propose an optimal hierarchical architecture for wireless IP-based network based on various analytical and simulation test results. The intention has been to develop an architecture, which would bring robustness as well as scalability to the current IP mobility scenario. We essentially tried to simulate different hierarchical scenarios by changing the mobility patterns in ns-2 [13]. In this context we define global users, who move around more than five subnets in the neighborhood of home subnet, showing greater mobility. Those with movements within five subnets are called local users. The purpose is to find out total network delay for each case as the number of subnets increases. We also varied the traffic handling capacity of agents at second, third and fourth... nth hierarchical models. Again for each case simulations were carried out with different mobility patterns separately generated in ns-2 of 10% global mobile users, 50% global mobile users and 90% global mobile users. We also found out the total signaling overhead suffered by the network and number of location update generated for each type of hierarchical networks. From the above results, we calculated the time-bandwidth product for different hierarchical architectures for 10%, 50% and 90% global users Analytical results For analytical calculation of various parameters of n- tiered architecture we considered the network of Fig.2. The network is assumed to have: f 1 number of L 1 FAs, i.e., each single subnet is controlled by a L 1 FA; f 2 number of L 2 FAs, i.e., f 2 = f 1 /x 1 ; f 3 number of L 3 FAs, i.e., f 3 = f 2 /x 2 ;... f n number of L n FAs, i.e., f n = f n-1 /x n-1 ; Performance gain will be more compelling if x 1, x 2, x 3...x n-1 are large and an MN spends a significant period of time within the domain controlled 12 China Communications December 2006

6 by a single L 2 FA, L 3 FA...RFA. The parameters chosen for analytical calculations have been shown in Table 1. Table. Parameters for analytical results Parameter Value Symbol Meaning f 1 Number of L 1 FAs) 100 f 2 Number of L 2 FAs 50 f 3 Number of L 3 FAs 10 f 4 Number of L 4 FAs 6 f 5 Number of L 5 FAs 4 f n Number of L n FAs 1 (RFAs)...up tosixth layer T 1, T 2, Registration times from 10 ms T 3, L 1 FA to L 2 FA, L 2 FA to (Typical T 4, T 5 L 3 FA,...L 5 FA to L 6 FA MAN delay) T 6 (T n ) Registration time from 200 ms L6FA to HA (Typical transpacific delay) l 1 Registration packet 46 bytes length from RFA to HA h 1, h 2, h 3, Number of hops between 2, 3, 4, h 4, h 5, h 6 L 1 FA and L 2 FA, L 2 FA 5, 6, 7 and L 3 FA,...L 6 FA and HA from one- and two-tiered architectures. The fall becomes more and more apparent as the number of MN increases. Looking at the present day scenario, one can assume that the number of mobile users will be increasing continuously. So in the near future, threetiered architecture should provide efficient performance with respect to LU. B. Signaling overhead Signaling overhead is a parameter that is of utmost importance for designing wireless network architecture. Fig. 4 shows the plot of total signaling overhead with number of tiers having subnet change period (ts) as the parameter. The plot shows that signaling overhead is enormously high for one-tiered architecture and goes low at two-tiered architecture slightly increases in threetiered architecture and steadily starts to increase from four-tiered architectures onwards. Since global signaling messages travel over a larger number of hops (and hence consume a larger portion of network resources), we have compared all the tiered architectures in terms of the total network capacity (aggregated over all hops) used in Fig. 5. A. Location update Fig. 3 shows the frequency of location updates generated by p MNs as they visit f 1 number of subnets one by one using one-tiered, twotiered,...n-tiered architectures respectively for a domain having 100 subnets. The figure shows that LU increases with the increase in the number of MNs. It also shows that LU tends to decrease with the increase in the number of tiers, which is obvious. But it points out a very important observation. It reveals LU comes to a saturation state at and from the third tier for any number of MNs after an abrupt fall Fig. Frequency of location update with number of tiers having MN as parameter for a -subnet domain China Communications December

7 architecture is only 40%. The percentage increase in signaling overhead for three-tiered over twotiered architecture is only 1.8% whereas the increase for four-tiered over three-tiered architecture is 10%. So with regard to location update and signaling overhead, three-tiered architecture is the favorable one. More results (Fig. 7 and Fig.8) will eventually prove the supremacy of three-tiered architecture. Fig. Total signaling overhead with number of tiers for a -subnet domain C. Handoff latency Table 2 shows a comparative chart of analytical estimate of handover delay for one-tiered, two-tiered... six-tiered From these plots it is clear that two-tiered architecture result in a significant reduction in the network-signaling overhead, especially when mobiles change subnets more frequently. As h n increases, the reduction in signaling overhead in two-tiered architecture becomes more significant. The plot also shows that the signaling overhead goes down, as the MN stays longer in a subnet (and domain) for all types of Fig. Signaling overhead with subnet change period aggregated over all hops tiered architectures. for different tiered architectures Thus this graph explains architectures. The values of T1, T2,...T5 are that in terms of total signaling overhead, two-tiered given in Table 1. The corresponding handoff architecture is favorable. However, looking at the location update plot in Fig. 3 one understands that latency plots with number of tiers have been three-tiered architecture proves to be advantageous. depicted in Fig. 6. The figure shows that the This is because the percentage decrease in LU from handoff latency remains constant at a high value two-tiered to three-tiered architectures is 80% for any level of FA change for single-tier whereas that between three-tiered and four-tiered architecture. For two-tier, the handoff latency is 14 China Communications December 2006

8 Table. Analytical estimate of handover delay Fig. Handoff latency with number of tiers very low so long the MN remains under the scope of same L 2 FA. For three-tier, the value is low so long the MN does not change its L 3 FA. Continuing in this fashion, we may notice that as we increase the number of tiers in a network, it produces low handoff latency so long the handoff is local, i.e., intradomain. For global handoff, the increase in tiers may cause a slight increase in handoff latency. This may be overshadowed if the number of location updates generated is low. Fig. 3 reveals that location update saturates at and from the three-tier. Also, Fig. 5 shows that signaling overhead is lowest at the two-tiered with a slight increase in three-tiered architecture. The signaling overhead increases sharply on the other two sides. So we may say that performance of three-tier architecture may prove to be favorable in terms of location update, signaling overhead and handoff latency. But we may be very sure about the optimality of three-tier architecture when we go through the simulation results in section Simulation Results A. Handoff latency The simulation result (ns-2) up to three-tiered architectures with regard to handoff latency is depicted in Table 3. Twotiered architecture showing a delay of ms is quite high as compared to ms delay of three-tiered architecture for regional handover. Handoff delay for three-tiered architecture improves by 89% over twotiered architecture and is very advantageous in large networks where an MN can change L 2 FAs under the scope of single L 1 FA very often. This would mean lower packet losses and faster China Communications December

9 handoffs on L 2 FA change. However, for local handover the delays of two-tiered and threetiered architectures remain the same. clear that three-tiered architecture is most favorable for IP-based wireless networks where the number of MNs making frequent global handoff is very high. B. Network delay 3.3 Discussions Fig. 7 shows the total network delay with hierarchical level for a network with 100 subnets for 10%, The various analytical and simulation results highlight the supremacy of three-tiered architecture. Table 50% and 90% global users. The figure clearly indicates that with the increase in the percentage of 4 shows the percentage change in value in different global users, delay increases for all the protocols. network parameters as we move from two- to fourtiered architecture via three-tiered architecture. Again it is found that network delay becomes saturated at and from the three-tier architecture. The The table shows that there is a drastic Table. A comparative chart of handover delay decrease in LU from two- to three- tier (ns -simulation results) (80%) but gradual fall from three- to fourtier (40%). Network delay shows an increase of 51% from two- to three-tier but only 8% fall from three- to four-tier. Threetier shows its limitation with respect to signaling overhead where there is an increase of 1.8% from two- to three-tier whereas a rise of 10% from three- to fourtier. The last parameter in the list shows an increase of 51% from two- to three-tier and an increase of 10% from three- to four-tier with respect percentage decrease in total network delay from twotier to three-tier is (~51%) and that between three- to delay-signaling overhead product. This parameter is an efficient way to determine the optimal result, tier to four-tier is (~8%). Since there is hardly any which shows that three-tiered architecture is the significant improvement beyond the third tier, so one optimal one. may understand that increasing number of tiers or So we may come to the conclusion that hierarchical architecture up to third level of hierarchy pro- C. Time-bandwidth product Fig.8 shows the plot of time-band- hierarchies beyond three is superfluous. width product with number of tiers. Looking at the plots one immediately understands the supremacy of three-tier architecture over all others. It shows that delay-signaling overhead (timebandwidth) product is lowest at three-tier with increase on either side revealing a U-shaped curve. It also highlights that this notch increases as the number of global Fig. Total Network delay vs. hierarchical level for -subnet domain users increases. So it becomes (ns- simulation result) 16 China Communications December 2006

10 locally the movement of an MN and disengaging the latter from frequent registrations with the HA. Like MIPv4 RR, TLMM uses hierarchy of FAs. However, in the former number of hierarchies of FAs to be used is not fixed. But in the latter number of hierarchies of FAs to be used is fixed at three. The L1FAs are called FAs, L 2 FAs are called Mobility Agents (MA) whereas L 3 FAs are known as Global Mobility Agents (GMA). The differ- Fig. Delay-signaling overhead product with number of tiers ent elements of TLMM as shown in Fig. 9 may de defined vides optimal performance in terms of vaious network parameters like location update, signaling overhead and handoff latency and delay-signaling overhead product. So we propose a three-tiered architecture called Three Level Mobility Model (TLMM) as the optimal hierarchical network architecture for IPbased wireless network. below. GMA: A GMA is placed at the boundary of two domains. It is a dynamically assigned Internet host in the MN s visited domain, which provides Global CoA for the MN. All incoming and outgoing packets are routed via the GMA. It can act as a proxy server capable of providing private address space. MA: An MA lying under the scope of a GMA is IV. OVERVIEW OF THREE LEVEL also an Internet host, dynamically assigned by MOBILITY MODEL the GMA on the MN s visited domain. It provides a 4.1 Architecture The TLMM uses three levels of hierarchy of FAs, located in visited domain, with the aim of handling more persistent Regional CoA (RCoA) to the MN. All incoming packets to the MN are routed by GMA through the MA. FA: A router on an MN s visited network, which provides configuration param- Table. A comparative chart of percentage change in value for differ- ent network parameters Percentage change in value eters like LCoA, RCoA and GCoA to the MN. DHCP server: A host that returns Parameter Two-tier Three-tier configuration parameters to to to the MN [14], [15]. In general, it Three-tier Four-tier assigns three CoAs to the MN: Location Update 80% decrease 40% decrease the LCoA, RCoA and GCoA. Network Delay 51% decrease 8% decrease MN: A host that changes its Signaling overhead 1.8% increase 10% increase point of attachment from one Delay-Signaling Overhead Product 51% decrease 10% increase subnet to another without China Communications December

11 changing its IP address. It is assigned three CoAs: local, regional and global by the network. HA: A router on an MN s home network (HN), which transfers datagrams to MNs that are away from HN and also maintains updated location information of the MN. 4.2 Mobility management mechanisms allow the use of private addressing and non-ip mobility management within the provider s own domain [16]. As soon as an MN enters into a subnet of TLMM, it sends a request for a new CoA to the FA. The FA in that subnet assigns an LCoA to the MN after authentication validation. The FA then sends proper route control messages to the MA. MA through Fig. Proposed three level mobility model The proposed TLMM takes care of macromobility (intradomain) and global mobility (interdomain) as shown in Fig. 9. A new node called GMA is introduced for this scheme at the network layer granularity higher than that of a subnet, thus reducing the generation of global location updates. By limiting intradomain location updates up to the GMA, the latency associated with intradomain mobility is reduced. The three level mobility management FA allocates a Regional CoA (RCoA) to the MN. A GCoA is then allocated to the MN via the selected route from GMA to MN through MA and FA handled by the GMA. In this way a path composed of sequential tunnels is established all along the way from GMA to FA through MA of the network to ultimately reach the address of MN under the scope of FA in the domain. It is the GCoA with which the MN registers with its HA 18 China Communications December 2006

12 via FA, MA and GMA. After registration with the HA, the mobility of the MN is handled locally, provided the latter moves within the boundaries of the visited domain under the same GFA. However, if the MN changes its GMA even within the same domain, it has to register with its HA so that the latter becomes aware of the new tunnel endpoint for packets addressed to the MN. The MN sends a Regional Registration Request message to the GMA through the MA and FA hierarchy for updating its location information within the domain. This message makes it all the way to GMA, if the MN s movement affects part of the path. In any case, the HA is not informed of the location-update procedure within the regional domain. Mobility management in TLMM can be categorized into two fields: Routing and Location Update. A. Routing TLMM uses the concept of IPv4 route optimization to [17], [18] eliminate the drawbacks of triangular routing [19]. Triangular routing is mainly a non optimal routing technique especially when the correspondent node is very close to the mobile node as it calls for greater delay and increased signaling overhead on the home agent. Route optimization is shown in Fig. 10. It deals with correspondent nodes maintaining an upto-date mobility binding of the mobile node s GCoA in their routing table. With this updated binding encapsulated packets may be sent from correspondent nodes to the mobile node s GCoA. Consequently the route optimization here is made up to the GMA level for higher security purpose. Packets may be sent by the CN to GMA of the foreign domain without interacting with HA after getting binding information about MN s GCoA from HA as shown in Fig. 10. It is the responsibility of the GMA to send the packets to the intended destination MN through MA and FA. In this method, however, the correspondent nodes need to be sure of the authenticity of the updates. Keeping this speculative feature in mind, the protocol has been designed to make the HA responsible for providing binding updates to any concerned correspondent node at foreign networks. In short there are four major steps involved for route optimization in TLMM. A binding warning control message sent to the HA indicates the unawareness of a CN about MN s CoA. Fig. Route optimization in TLMM China Communications December

13 A binding request sent by CN. A binding update sent by HA containing MN s current GCoA. A binding acknowledgement sent by CNs to HA for smooth handoffs. Packets may be sent by the CN to GMA of the foreign domain without interacting with HA. It is the responsibility of the GMA to send the packets to the intended destination MN through MA and FA. TLMM is advantageous in large networks where an MN can change MAs very often. This would mean lower packet losses and faster handoffs on MA change. However, when FA change occurs, the delays of two-tiered architecture and TLMM remain the same. registration request and registration reply messages sent and received by GMA (Figs. 11a and 11b). TLMM registration request message consists of the following fields. Similar packets sent and received by MA and FA should contain RCoA and LCoA instead of GCoA and HA should be replaced by GMA and MA respectively. The registration request/reply messages consist of the following fields: Type: 1 (registration request), 3 (registration reply) S: Simultaneous bindings B: Broadcast datagrams D: Decapsulation by MN M: Indicates that the HA should use minimal encapsulation. B. Location update and registration V: Indicates that the HA should use Van Jacobean header compression. The MN performs home registration when it is G: Indicates that the HA should use GRE assigned a new GCoA after entering a new visited domain. It transmits the TLMM registration request encapsulation. packet (Fig. 11a) to the HA via GMA and sets the Lifetime: The number of seconds before the CoA field to GCoA. This results in the establishment registration is considered expired. of a tunnel between the HA and the GMA for routing Home agent: The IP address of the MN s home agent of packets addressed to the MN s home address. The FA and corresponding MA should also process the registration message to establish the path between GMA and MA and another between MA and FA. So the message is not sent directly to GMA but is processed by MA and FA as well. TLMM registration involves four steps. The MN requests the forwarding service by sending a registration request to the GMA via FA and MA that the MN wants to use. Home address: The permanent IP address of the MN in the home network GCoA: The IP address of the corresponding GMA RCoA: The IP address of the corresponding MA LCoA: The IP address of the corresponding FA Identification: A 64-bit number generated by the MN, used for matching registration requests to registration reply and for security purposes Extensions: Authentication extension Code: Indicates result of the registration request The GMA relays this request to MN s HA. Bit The HA either accepts or rejects the request and Type S Lifetime B D M V G rsv sends a registration reply to the GMA. GMA relays this reply to the MN via MA and FA. The TLMM registration operation uses two types of messages, carried in UDP segments. These are TLMM Home agent Home address Global care-of-address (GCoA) Regional care-of-address (RCoA) Local care-of-address (LCoA) Identification Extensions Fig. a TLMM registration request message 20 China Communications December 2006

14 The registration request message from MN is directed towards the corresponding FA. The FA removes the LCoA field and directs it towards the corresponding MA. The MA in a similar manner removes the RCoA field and sends the message to GMA. The GMA send the message to HA. So in this way FA maintains the LCoA- MN identification link, MA maintains the RCoA-LCoA link, and GMA maintains the GCoA-RCoA link whereas HA maintains the home address-gcoa link of the MN in the foreign network. The registration reply message from HA follows the similar path from HA to MN via the intermediated agents of the network. Bit Type Code Lifetime Home agent 4.4 Architectural comparison Home address Global care-of-address (GCoA) Regional care-of-address (RCoA) Local care-of-address (LCoA) Identification Extensions Fig. b TLMM registration reply message Functionality of TLMM differs from the two hierarchical architectures, viz., HMIPv6 and MIPv4 RR. Like MIPv4 RR, TLMM packets are actually addressed to GMA and then tunneled within the access network in TLMM. However in HMIPv6, packets are destined to an address within the MAP s subnet and not to the MAP itself, and so the latter has to function as a typical HA for intercepting data packets and tunneling them to the MN s location. An MN in a TLMM domain is assigned three CoAs: an LCoA of FA, an RCoA of MA and a GCoA of GMA. However it is assigned two IP addresses in HMIPv6: an RCoA on the MAP s subnet and an on-link CoA (LCoA) that corresponds to the actual location of the MN in a MAP domain. In MIPv4 RR the number of CoAs depends on the number of layers of FAs. V. CONCLUSION In a growing world of today where the number of mobile users is increasing everyday, networks of tomorrow will be large in size. But as the network size increases, the problem of scalability creeps in. This proposed model is an efficient solution to the scalability problem. TLMM shows its efficiency when the number of subnets increases, providing lower network delays. Even with a higher number of users with greater mobility, TLMM stands as a robust model providing lower network delays and far-reaching improvement in handoff latency for intradomain changes. The global signaling overhead up to the HA is also reduced enormously. TLMM has a total signaling overhead that is comparable to that of a two-tiered architecture and moreover it has very low handoff latency for intradomain changes as well. The effective location updates are also lesser compared to other tiered architectures architectures. The supremacy of TLMM over other tiered architectures lies in its lowest time-bandwidth product for different types of global users. It is a clear indication of the fact that with increased number of global location updates and handoff and with ever-increasing number of MNs in the present day scenario performance of TLMM is the most optimal one. So it is jusfied to establish TLMM as an optimum model for next generation IP-based networks. One of the open issues that challenge us is the integration of wireless quality of service (QoS) among end users in TLMM. Work is done on such an endto-end QoS support. VI. REFERENCES [1] C. Perkins, IP Mobility Support, IETF RFC China Communications December

15 2002, Oct [2] D. B. Johnson, hierarchical Foreign Agents and Regional Registration, Minutes of the MIPv4 Working Group Meeting, IETF, Mar [3] S. Deering and R. Hinden, Internet Protocol Version 6 (IPv6) Specification, IETF RFC 2460, Dec [4] D. Johnson and C. Perkins, Mobility Support in IPv6, IETF draft, draft-ietf-mobileip-ipv6-15.txt, July [5] A. T. Campbell et al., Comparison of IP Micromobility Protocols, IEEE Wireless Commun., Feb. 2002, pp [6] Salkintzis, Mobile Internet- Enabling Technologies and Services, CRC press. [7] H. Sollman et al., Hierarchical MIPv6 Mobility Management, IETF draft, draft-ietf-mobileiphmipv6-05.txt, July [8] J. Manner et al., Mobility Related Terminology, IETF draft, draft-manner-seamobyterms-01.txt, Mar [9] M. Chakraborty, I.S.Misra, D.Saha, A. Mukherjee, A Comparative Study of Existing Protocols Supporting IP Mobility, International Journal of Information and Computing Science (IJICS), Vol. 5, No. 2, Dec 2002, pp [10] D. Saha, A. Mukherjee, I. S. Misra, M. Chakraborty Mobility Support in IP: A Survey of Related Protocols, IEEE Network Magazine, Vol. 18, No. 6, November/December 2004, pp [11] S. Das, A. Misra, P. Agrawal, and S.l K. Das, TeleMIP: Telecommunications-Enhanced MIPv4 Architecture for fast Intradomain Mobility, IEEE Personal Communications, Aug 2000, pp [12] S.Das, A.McAuley, A.Dutta, A.Mishra, K. Chakraborty, S.K.Das, IDMP: An intradomain mobility management protocol for next generation wireless networks, IEEE Wireless Communications, June 2002,pp [13] ns-2 home page, nsnam/ns. [14] R. Droms, Dynamic Host Configuration Protocol, IETF RFC 2131, Mar [15] J. Bound et al., Dynamic Host Configuration Protocol for IPv6 (DHCPv6), IETF draft, draftietf-dhc-dhcpv6-24.txt, Oct [16] I. S. Misra, M. Chakraborty, D. Saha, A. Mukherjee, A Three Level Hierarchical Mobility Management Architecture for IPv4 Networks, Proc. International Conference IST Mobile and Wireless Summit 2004, pp , June 2004, Lyon, France. [17] C. Perkins and D. B. Johnson, Route Optimization in Mobile IPv4, Internet draft, draft-ietfmobileip-optim-07.txt, [18] C. Perkins and D. Johnson, Route Optimization in Mobile IPv4, Internet draft, draft-ietfmobileip-optim-08.tyxt, Nov [19] I. Saha Misra, M. Chakraborty, D. Saha, A. Mukherjee, An Approach to Optimal An Approach for Optimal Hierarchical Mobility Management Network Architecture, Proc. IEEE VTC-2006, May, Melbourne, Australia. BIOGRAPHIES Mohuya Chakraborty (mohuyacb@yahoo.com) presently holds the post of Asst. Professor in the Department of Electronics and Communication Engineering, Netaji Subhash Engineering College, Kolkata, India. She received her B.Tech degree in radio physics and Electronics from Calcutta Univerity (1994) and her M.Tech. degree in radiophysics and electronics from Calcutta University (2000). She is presently pursuing her Ph.D. in the field of mobile communication and networking at Jadavpur University. Her areas of interest include mobility management and QoS support in IP-based wireless networks. Her other research activities are related to VLSI design with FPGA. She has authored some journals and international conference papers. 22 China Communications December 2006

16 Iti Saha Misra in) presently holds the post of reader in the Department of Electronics and Telecomm u n i c a t i o n Engineering, Jadavpur University. She received her B.Tech. degree in radiophysics and electronics from Calcutta University (1989) and her Master s in telecommunication engineering from Jadavpur University (1991). She completed her Ph. D. in engineering in the field of microstrip antennas at Jadavpur University (1996). Her current research interests are in the areas of mobility management network architecture and protocols, integration architecture of WLAN and 3G networks, and location management for cellular wireless networks. Her other research activities are related to microstrip antennas and design optimization of wire antennas using numerical techniques. She has authored several journal and international conference papers. She is the recipient of the prestigious Career Award for Young Teachers from the All India Council for Technical Education (AICTE) for financial year She is the founder chair of IEEE Women in Engineering, Calcutta Chapter. Debashis Saha (ds@iimcal.ac.in) is currently a professor in the MIS and Computer Science Group of the Indian Institute of Management Calcutta (IIMC), India. He received his Bachelor s degree from Jadavpur University, and his Master s and Ph.D. degrees both from Indian Institute of Technology (IIT), Kharagpur, all in electronics and communications engineering. His present research interests include pervasive communication and computing, wireless networking and mobile computing, and WDM optical networking. He has published more than 120 papers in various conferences and journals, and directed four funded projects on networking. He has co-authored a monograph and five books, including Networking Infrastructure for Pervasive Computing: Enabling Technologies and Systems (Kluwer, 2002) and Location Management and Routing in Mobile Wireless Networks (Artech House, 2003). He is the recipient of the prestigious Career Award for Young Teachers from AICTE, Government of India, and is a SERC Visiting Fellow, Department of Science and Technology (DST), Government of India. Amitava Mukherjee (amitava.mukherjee@in.ibm. com) is a senior consultant at IBM Global Services, Calcutta. Currently, he is a senior researcher at LCN/IMIT, Royal Institute of Technology, Sweden. His research interests are in mobile computing and wireless communication, pervasive computing and mobile commerce, optical networks, combinatorial optimization, and distributed systems. He received a Ph.D. in computer science and engineering from Jadavpur University. China Communications December