Telecommun Syst DOI 10.1007/s11235-009-9169-6
Routing optimization over network mobility with distributed home agents as the cross layer consideration Chung-Sheng Li · Fred Lin · Han-Chieh Chao
© Springer Science+Business Media, LLC 2009
Abstract Routing optimization (RO) is an important support for Mobile IPv6 while performing handover. It can reduce the overhead of a home agent and make the traffic more fluent between MN and CN; however, Network Mobility (NEMO) in Mobile IPv6 lacks the capability of RO. In this paper, we propose an enhanced architecture by deploying Home Agents Location Registration Agents (HALRA) on cellular data switch (3G/B3G) network incorporating with Distributed Home Agents (NEMO-DHA) on NEMO for a Mobility Critical Area. Under this architecture, it can provide less handoff latency on the Mobility Critical Area and make a detour to avoid the perplexed NEMO routing optimization problem. Moreover, NEMODHA can also solve a sort of Nested-NEMO problems by setting authorized Mobile Routers to serve as a set of Distributed Home Agent, and we provides a Cross-Layer suggestion by utilizing R-bit to schedule fairly for NEMO connections. The wireless network can also benefit from Distributed Home Agents by cooperating cellular data switch network with sharing HALR resource. Most models in the paper are described through UML modeling language evaluated by a mathematical model and simulated by NS2Mobiwan2 (http://www.ti-wmc.nl/, 2009). C.-S. Li () · F. Lin Department of Electrical Engineering, National Dong Hwa University, Hualien, Taiwan e-mail:
[email protected] F. Lin e-mail:
[email protected] H.-C. Chao Department of Electronic Engineering, National Ilan University, I-Lan, Taiwan e-mail:
[email protected]
Keywords Mobile IPv6 · NEMO · Home agent · NS2
1 Introduction The recent advances in portable devices and wireless networks have resulted in a new paradigm of computing to access Internet while users changing their positions frequently. Mobility support for IPv6 (Mobile IPv6, MIPv6) [1] is one of the most important technologies when the continuous connectivity of mobility environment is taken into consideration. There are some strong demand scenarios of IP mobility in the real world, such as passengers in MRT or city buses, etc. If a mount of mobile devices request to change their attachment at the same period of time, a handoff bottleneck will be encountered. Obviously, MIPv6 is not good enough for group mobility, because the original design of MIPv6 is only for a single device. Network Mobility (NEMO) proposed in RFC 3963 [2] is expected to scale for a potentially large number of devices in mobile networks. It provides the basis support of group mobility and is adopted by several important vendors, such as Cisco. There are also some open or commercial sources available for Linux [3], FreeBSD [4], and Cross Platform [5]. In the NEMO Basic Support protocol compliant mobile network, Home Agent (HA) is the only anchor point for Correspondent Node (CN) and Mobile Node Network (MNN) [6]. It lacks support of RO and makes a longer latency in HA. In our experience of setting up a MIPv6 testbed with MIPL [3] and Futuresoft [7], the configuration of Home Agent is not harder than ones of other network services. We believe this reason causes Multiple Home Agents architecture with potential to be prevailing for the coming network environment.
C.-S. Li et al.
In this paper, we extend NEMO basic support with Distributed Home Agents (DHA). A Home Agent Location Register (HALR) is used to maintain the Home Agents location information and HALRAs play the role to redirect ICMPv6 DHAAD request on demand. To deploy HALRAs will reduce the handoff latency for a busy core network. The mechanism also makes the support of authorization, dynamic Home Agent assignment, and inter-network mobility support possible. After analysis of our setup scenarios, the NEMO-DHA architecture also has the potential to decrease handoff latency between cellular data switch (3G/B3G) network and Internet, no matter it is in a tightly coupled or loosely coupled inter-working hybrid network. The rest of the paper is organized as follows. Section 2.1 gives the introduction of Network Mobility, Sect. 2.2 shows what are the problems of the NEMO Routing Optimization (NEMO RO) problems; Sect. 2.3 gives the introduction of the distributed home agents (DHA). Our NEMO-DHA architecture and MR-DHAD mechanism are present in Sect. 3. Internetworking models of the hybrid network are present in Sect. 4. Simulation results, mathematical analysis and discussions are presented in Sect. 5. Finally, the conclusion is remarked in Sect. 6.
2 Related works 2.1 Network mobility NEMO Basic Support is a solution to preserve session continuity by means of bi-directional tunneling between MRs and their HAs much like what is done with Mobile IPv6 [17] for mobile nodes when Routing Optimization is not used. In the case of normal fixed network, data packets are sent to the destination directly. But in the case of Mobile IP and NEMO Basic Support protocol, all of the packets forwarded by Mobile Node (MN)/Mobile Router (MR) in the outbound Fig. 1 Illustration of mobile router
direction have to go through HA first. NEMO Basic Support has become IETF proposed standard in Jan. 2005. This protocol allows session continuity for every node in the mobile network as the network moves. Every node in the mobile network is reachable while this network is changing its position. NEMO Basic Support protocol is backward compatible with MIPv6, so HA can operate as a MIPv6 HA as well. In addition, NEMO Basic Support protocol defines required operations and messages for basic group IP mobility support, and let terminal devices access multiple networks transparently through the MR. A MR is located at the edge of the MNN and connects the MNN to the backbone of Internet. Consequently, a MR can be assigned as a default gateway for the MNN, which can include both fixed and MNs behind the MR; then, the Internal network topology of MR keeps relatively stable when the MNN is migrating. In NEMO Basic Support, only the MR and the HA are NEMO-enabled. The following will describe them briefly. 2.1.1 Mobile router (MR) A MR is capable of changing its point of attachment to the network, forwarding packets between two or more interfaces, and possibly running a dynamic routing protocol. As shown in Fig. 1, a MR acts as a gateway between an entire MNN and the rest of Internet, and has one or more egress interfaces and one or more ingress interfaces. Packets forwarded upstream to the rest of the Internet are transmitted through one of the MR’s egress interface; packets forwarded downstream to the MNN are transmitted through one of the MR’s ingress interface. NEMO-enabled home agent (NEMO-HA) In Fig. 2, it shows the NEMO-enhanced ICMPv6 Dynamic Home Agent Address Discovery (DHAAD) Request message, which adds one more Mobile Router Support Flag (R).
Routing optimization over network mobility with distributed home agents as the cross layer consideration
Fig. 2 Dynamic home agent address discovery request message with NEMO support Table 1 Modified fields of dynamic home agent address discovery request message for NEMO basic support Packet field
Field
Default
ID
size
value
Type
8 bits
144
Code
8 bits
0
Note
Checksum
16 bits
The ICMP checksum
Identifier
16 bits
Aid in matching Reply messages
R
1 bit
0
1 if NEMO support
Reserved
15 bits
0
Unused
to this Request message
If a NEMO-HA receives a DHAAD request message with the Flag set, it MUST reply with a list of NEMO-HAs supporting MRs. If none of the NEMO-HAs support MRs, the NEMO-HA MAY reply with a list of HAs that only support Mobile IPv6 MNs. The modified message format and field details (Table 1) are as follows. As described above, most of the NEMO operations are inherited from MIPv6, so a HA may respond for multiple MNs on the home link. Nevertheless, the failure of a single HA can result in the loss of connectivity for many MRs located throughout the Internet, which would cause more serious problems than MIPv6. 2.2 NEMO routing optimization problems As IPv6 provides the world of global IP addresses, P2P connection between MNs and CNs in MIPv6 introduces the mobility Routing Optimization (RO) features and problems. In addition, NEMO, which the network itself is moving entirely, extended the MIPv6 protocol and intensified the RO problems. The protocol hides the MR mobility by making as if the MR was always connected to a Home Link. With current NEMO Basic Support, all communications to and from MNNs must go through the MR-HA tunnel when the mobile network is away. Therefore, all the packets MR-CN flow should be via the single HA without route optimization. This approach insures the packet will be delivered to proper MNN but also results in longer packet route and larger packet delay. In short, the longer route leads to the increased delay, packets overhead, and chances of packet frag-
mentation. Because of nested NEMO network where multiple levels of mobility are formed, NEMO RO problems have more challenges and more difficulties than MIPv6 RO problems. Compared to MIPv6 which provides ‘Return Routability (RR) mechanism as a standard solution for routing optimization, the NEMO RFC seems to describe less about RO problems. Although MR can use RO just like any Mobile Host and the NEMO home agent allow route optimization to mobile network nodes known to be attached to a mobile router, the Local Fixed Node which is hosted by Mobile Router does not implement RO because it’s not a MIPv6 node. There are a couple of drafts such as draft-Thubert [8], draft-Zhao [9], draft-Watarti [10], and draft-Clausen [11], etc. and a lot of discussions on the NEMO working group mailing list, which address and analyze those NEMO Route Optimization Problems. Thubert [8] coined three user cases to illustrate the reasons why any sort of RO in NEMO is desirable: 1. You have a high quantity of traffic and transmitting all these packets via the home agent would consume too many network resources if the CN is relatively closer to the MNN on the direct path. 2. You have a small quantity of traffic, but it is very sensitive to delay, so you need the shortest path if the CN is relatively closer to the MNN on the direct path. 3. Each radio hop adds a probability of loss. If the traffic becomes useless at a given degree of loss, RO might be highly desirable. Nevertheless, in order to achieve the optimal route in NEMO, MR and Corespondent Agent (CA, which denotes either CN or Correspondent Router (CR)) should become anchor points too. Instead of basic MR-HA bi-directional tunnel, there are two kinds of RO problems remaining to be solved. The first one is MR-CR tunnel; in specific, MR will search the nearest CR to CN, and make a binding with it. CR is binding some kind of anycast address to ensure the discovery process possible. The binding process between MR and CR are activated to acquire some required information such as MR, CR prefixes. One example of that is MNN to MNN RO, which might save a number of radio hops in a (fixed or mobile) mesh network. Another scenario is named as “pinball” or “dog-leg” routing [8], which figuratively illustrates a tortuous path for the case that multiple mobile networks are nested. It incurs very inefficient routing depending on the relative location of HAs however, extra IP header for IP tunneling is added per level of nesting to all the packets. 2.3 Distributed home agent with distributed home agent discovery mechanism In the current specification of MIPv6, a MN can have one single HA on the home link during the connection, so the
C.-S. Li et al.
overhead can not be ignored if a MR moves far away from a HA. Another, less typical example is if the home link is not reachable by some reasons, the MN loses its connections with CNs which do not keep the binding information. At this time, a MN cannot initiate a new connection even if the foreign link is working without any trouble. Consequently, the distributed home agent architectures (DHA) were motivated to solve those problems. Zhao [9] describes that deploying multiple HAs in the different domains or spreading binding information needs to consider a lot of issues, such as fundamental changes, conflict, and scalability. Several approaches to make MR dynamically switches to the topologically closest HA are considered [12]. It is noteworthy that Kyoung [13], Esaki [14] shows the possibility to use multiple-HAs through routing protocol, but the limitation is its deployment flexibility. In the research, the proposal of Yen [15] choose using anycast address on MN, and deploy few functions on Border routers, it also shows the DHA is possible on MIPv6 network. However, the question is currently MN can’t use anycast IP in the requirement of IPv6 anycast IP delegation rules [15]. In our case if the NEMO-DHA is based on anycast IP, NEMO-MR is one kind of router, and it still fits the IP delegation rules [16].
3 NEMO-DHA
2.4 Cross layer suggestion for NEMO in DHA
We propose to deploy DHA on MCA, which indicates those areas with the most numbers of handoff activities; therefore, well-deploying DHA on top or inside of MCA is the key of achieving meaningful approximately RO and effective investment because NEMO-DHA only affects MRs and HAs. In sum, these characteristics make the network more stable by preventing MNs from resetting message overflow. In Fig. 4, it shows the relationship of MCA by UML User Case diagram.
As shown in Fig. 3, since a group of users can connect to Internet through the same MR, it is necessary for DHA to allocate more bandwidth to provide quality of service than a single user’s connection. Although some literary works, such as [23–25], are for Mobile IP, their efforts are mainly on reducing of handover time. However, they didn’t take NEMO into consideration. In our NEMO DHA architecture, we suggest to use R-bit to identify a NEMO connection and allocate more bandwidth or give a higher priority to NEMO queues, since NEMO connections act as a backbone between DHA and MR. For achieving this goal, DHA should have a fair scheduling for each wireless queue. In [22], it provides some cross-layer mechanisms, which can schedule the priority queue of each connection.
Fig. 3 DHA queue scheduler for NEMO in DHA
By digesting the advantages of those proposals, we try to make the DHA approximate-RO be the solution over NEMO. We form NEMO-DHA by introducing HALR in network, and use HALRAs with MR-DHAD mechanism to respond the nearest HA and get the approximately optimized MN-HA-CN routing route. Without a proper RO solution, the network performance becomes worse especially in nested NEMO. It is easy to know that RO helps to get the best transmission performance but several pre-configuration procedures should be completed in advance, so its cost is severe transmission delay. In some circumstances, entirely RO seems not to be needed. Based on these kinds of networks, our proposal will only affect the area which deployed HALRAs, and make no change to other area. It allows service providers to have great deployment flexibility on deciding which mobility area they need to enhance. We recommend deploying DHA on Mobility Critical Area (MCA) with the highest priority. The obvious example of MCA is driving along the super highway, since there are a lot of moving vehicles equipped with MRs in the future. 3.1 Consideration of deploying distributed home agents
3.2 Setup the home agents location register (HALR) Contrast with pure multiple HAs, which may suffer from multiple HA’s scalability or Border Router’s (BR) rout-
Fig. 4 Relationship of mobility critical area
Routing optimization over network mobility with distributed home agents as the cross layer consideration
Fig. 5 Deploy HALR in cellular data switch network
ing table explosion, we propose to use centralized Home Agents Location Registration (HALR) in cellular data switch (3G/B3G) network to adjust the HAs’ Routing exchange and avoid un-hitting HA penalty. As shown in Fig. 5, HALR is settled in cellular data switch (3G/B3G) network to enhance the mobility support. Liking Home Location Registration (HLR), which maintains the mobile Home location in cellular network, HALR will attach to the SGSN and maintain the pre-registered HAs list. A HALRA will respond with HA addresses when a MR issues a HAs request. Deploying HALR makes DHAs become manageable in centric management center and provide a trusted certifier to authorize and certificate the DHAs. By incorporating HALR, with some extra effort, the rapid fault detection and fault tolerance of DHAs become possible. In Fig. 5, we neglect the connection from SGSN and HLR to Inter-working Function Unit because this connection is basically to provide the correspondent Internet cellular support.
Fig. 6 Sequence chart of transmission phase
3.4 MRDHAD mechanism Mobile Router Dynamic Home Agent Discovery mechanism (MR-DHAD) takes the advantage of NEMO’s MIPv6extended DHAAD request messages. As shown in Fig. 2 which has been described in RFC 3963. A new Flag (R) is introduced in the DHAAD Request message. If a HA receives a DHAAD request message (DHAAD Req.) with the Mobile Router Support Flag set, it must reply with a list of HAs that support MRs. Our proposal is to embed HALRAs in router’s IOS function or as a 3G/B3G RNC’s module over network. HALRAs can judge whether the quest should be processed by detecting “R” bit, which is defined in NEMO Basic Support RFC. The full MR-DHAD mechanism can be classified into three phases, which are transmission phase, registration phase, and maintenance phase. 3.4.1 Transmission phase
3.3 Deploying home agents location registration agents Home Agents Location Registration Agents (HALR Agents, HALRA) could be deployed in BR’s IOS function or as a RNC module, it’s a small function set to recognize R-bit of NEMO Basic Support in MIPv6 DHHAD messages and help MRs to acquire the nearest HA with the MR-DHAD mechanism. The behavior of HALRAs is similar to the mechanism in DNS; that is, it can receive the domain name (Anycast HA address) and reply requests with the nearest HA’s unicast address. For a managed network, each authorized router has the capability to be activated as a HA, and each node in backbone could carry the HALRA function. Because currently the authorized DHA is not popular to network, most of HALRAs likely reserve only one or two nearest DHAs to reply the home agent discovery request messages.
In transmission phase, as shown in Fig. 6, when a MR is activated in MCA, it sends the DHAAD request to acquire the HA. When the message goes through any router with HALRA function support, this router will identify if the Rbit is set (which means the requester is MR), the HALRAs function will reply DHAAD request to the nearest HA immediately if the nearest HAs record are already in its short HA location list. Otherwise, it will perform several Home Agent Request with HALR to receive near/proper HAs list. The procedure of transmission phase as shown in Fig. 7 will shorten the routing path greatly, but such a case will be activated rarely since most of MRs support R-bit. 3.4.2 Registration phase The registration phase as shown in Fig. 8 are used while deploying new HALRA. Specifically, while a MR sends a
C.-S. Li et al.
DHAAD request to HALRA and HALRA has no HAs list, HALRA will send HA request to ask HALR for HAs list, which is replied by a HA Ack message.
In both Inter-working model, the WLAN service provider could register their HAs on cooperating network’s HALR to provide less handoff latency by DHAs.
3.4.3 Maintenance phase
4.1 Tightly coupled inter-working
Maintenance phase as shown in Fig. 9 are used to activate the HAs list renew procedures. HALR will send HAadv with new HAs list to HALRA; then, HALRA will update its HAs list to the last status. MR-DHAD mechanism doesn’t contain the HA cooperation mechanisms, but it provides search and dynamically deployment ability on NEMO network.
In tightly coupled internetworking, HALRAs could be deployed directly in cooperating network as in Fig. 10.
4 Hybrid LAN integration model Extending from public hotspots, such as airports and hotels, to public transit vehicles, such as bus, railway, MRT, the main idea of how the cooperating network benefits from proposed cellular data switch network with sharing HALR is the policy of HALRA deployment.
4.2 Loosely coupled inter-working In loosely coupled internetworking, as shown in Fig. 11, HALRAs can’t be deployed directly in cooperating network. After analyzing the cooperating scheme, we would attach the HALRAs on WLAN (which assumed as our cooperating network) Gateway, which does not have any direct link to cellular data switch network elements, but it uses AAA services to inter-work with the cellular data switch network’s home network AAA servers.
5 Evaluation 5.1 Handoff delay model of NEMO Stefano extended the performance analysis of Mobile IP and used it to compare with Route Optimization as in MIPv6 [18]. Based on Stefano’s analysis of MIPv6 RO Handoff Delay (HOD) model, we apply it to the analysis of a Handoff Delay Model with NEMO Basic support for Egress Interface (which does not take the inner MR case (or MNN into consideration). Figure 12 shows the Handoff Delay Model of NEMO Basic support for Egress Interface. The line distance between nodes is measured in number of hops.
Fig. 7 Sequence chart of transmission phase exception
Fig. 8 Sequence chart of registration phase
k: denotes number of hops on shortest path between BR and HA l: denotes number of hops on shortest path between BR and CA
Routing optimization over network mobility with distributed home agents as the cross layer consideration
Fig. 12 Handoff delay model of NEMO basic support for egress interface
tion delay (REGDHA ). So in NEMO Basic support, the REGDNEMO = REGDHA is expressed as: Fig. 9 Sequence chart of maintenance phase
• Registration delay (REGDHA ) REGDHA (k) = Cp,HA + 2(k + p)Ctx .
(1)
The normal one way transmission latency of NEMO Basic Support (TLNEMO ) will be: TLNEMO = Cp,BR + Cp,HA + Cp,CA + (m + k + p)Ctx . (2)
Fig. 10 Deploy HALR in tightly coupled network
• Transmission latency (TLNEMO ) If we have several competitor mechanisms to compare, the delta function is introduced to verify the handoff delay and transmission latency. • Registration delay: REGD = REGD1 − REGD2 .
(3)
Transmission latency TL = TL1 − TL2 .
(4)
5.2 Handoff delay model of NEMO-DHA In our NEMO-DHA scheme, the Handoff Delay Model of NEMO-DHA could be modeled as illustrated in Fig. 13. Fig. 11 Deploy HALR in loosely coupled network
m: denotes number of hops on shortest path between HA and CA p: denotes number of hops on shortest path between BR and MR Cp : denotes cost to process one message at a node Ctx : denotes cost to transmit one message over one hop. The practical approach to evaluate those system implementations is to use “traceroute” to obtain above hop numbers and time delay. In NEMO Basic support, the registration delay (REGDNEMO ) without RR procedure has only Registra-
q: denotes number of hops on shortest path between BR (with HALRA function) and the nearest DHA, which q ≤ k. s: denotes number of hops on shortest path between DHA and CA (further discussion will be presented in Analysis by NEMO CA placement section). r: denotes number of hops on the shortest path between DHA and original HA. HALRA is deploying in the path between MR to BR. It will reply the DHAAD request to the nearest DHA. If the request R-bit is not set, the number of hops on shortest path between BR and nearest DHA q is equal to the number of hops on shortest path between BR and HA k. The registration delay (REGDNEMO-DHA ) equation will be similar to REGDNEMO . It also has three components:
C.-S. Li et al. Fig. 14 NEMO foreign CN case
Fig. 13 Handoff delay model of NEMO-DHA
Fig. 15 NEMO-DHA foreign CN case
• Registration delay (REGDDHA ) REGDDHA (q) = Cp,DHA + 2(q + p)Ctx .
(5)
• Authorization and Certification procedure delay (ACD) ACD(r) = Cp,DHA + Cp,HA + 2rCtx .
(6)
This equation is varied for different DHA cooperating mechanisms are adapted. And the normal one way transmission latency of NEMO-DHA (TLNEMO-DHA ) will be: • Transmission latency (TLNEMO-DHA ) TLNEMO-DHA = Cp,BR + Cp,DHA + Cp,CA + (s + q + p)Ctx .
(7)
In Approximate Routing Optimized NEMO-DHA scheme, we can set the distance q to be a number between 0 and k (0 ≤ q ≤ k) since the BR could act as the DHA (q = 0) or DHA could just be the normal HA (q = k). The number of hops l on shortest path between BR and CA will be equal to the number of hops s on shortest path between DHA and CA plus the number of hops q on shortest path between BR and nearest DHA. The original transmission latency could be simplified as: • transmission latency (TLNEMO-DHA ) TLNEMO-DHA = Cp,DHA + Cp,CA + (s + p)Ctx .
(8)
Another Approximate Routing Optimized case is MR to be deployed as another DHA; that is, MR is also integrated into DHA. Consequently, we can get l = s = s + p, and r = r + p. Then, • transmission latency (TLNEMO-DHA ) TLNEMO-DHA = Cp,BR + Cp,CA + Cp,DHA + s Ctx .
to MR, it means that the routing path is shorter than the case of a single HA. It can be denoted as q ≤ k. Thus, the value of NEMO-DHA Registration delay (REGDNEMO-DHA ) is always shorter than the one of REGDNEMO schemes at a distance s.
(9)
To compare the NEMO-DHA with the normal NEMO scheme, the difference of the REGD value of NEMO and NEMO-DHA is variable k and variable q. If the HA is closer
5.3 Analysis by NEMO CA placement For comparing the transmission latency, the parameter of shortest path between DHA and CA, s, should be carefully addressed. We could roughly classify NEMO CA placement scenario as outer BR and inner BR cases. 5.3.1 CA outter home domain case (COHD) The CA Outter Home Domain case could be classified into two types: Foreign CN case and Neighbor CA case. Foreign CN (FCN) case: In most cases, the CAs are available in foreign domain, so basically in NEMO-DHA, we assume that q is always shorter than s. It can be depicted reasonably that the average CA access in HA-CA hop numbers will be converged approximately to a constant number. While q ≤ k, NEMO-DHA will be better than NEMO in the Foreign CN case. For extended NEMO-DHA scheme (cooperated with other tunneling-RO, i.e. RRH [19] or ARO [20]), FCN case can be turned into the CA inner Home Domain case. Neighbor CA (NCN) case: In NEMO, if CA is placed in Home domain, we call it a Neighbor CA. For Neighbor CA
Routing optimization over network mobility with distributed home agents as the cross layer consideration Fig. 16 NEMO neighbor CA 1st case
Fig. 19 NEMO-DHA neighbor CA 2nd case
Fig. 17 NEMO-DHA neighbor CA 1st case
Fig. 20 NEMO CA inner home domain case
Fig. 18 NEMO neighbor CA 2nd case Fig. 21 NEMO-DHA CA inner home domain case
case, the CA is inside or near the BR. Variable s can be expressed as: s = q + w, (inside BR) . (10) s = q + w + λ (near BR)
Fig. 22 NEMO CA inner MR case
w denotes the hop numbers between BR and MR while CA is placed between BR and MR (not restrict to the shortest path, but w usually smaller than p). λ denotes the hop numbers between BR and CA. This case is the most difficult one to be analyzed. In NCA case, if CA is placed inside BR, for the Approximately Routing Optimized NEMO-DHA scheme, it could be turned to CA inner Home Domain case. 5.3.2 CA inner home domain case (CIHD) For CA Inner Home Domain case which the CA is inside the BR, the variable s can be expressed as: s = q + w + γ.
(11)
w denotes the hop numbers between BR and MR while CA is placed between BR and MR (Not restrict to the shortest
path). γ is the number of hops if CA is placed inside the MR. In CIB, NEMO-DHA could help the system to span the action area of MRHA tunneling consumption in q. The distance of HA to MR q is proportional to the system throughput and packet delay variance. In most cases, although Approximate RO has no significant delay or latency optimization, it consistently improves the overall system delay and throughput toward the best case solution. It’s a adaptive solution for most NEMO connections.
C.-S. Li et al.
5.4 Simulation The MR-DHAD mechanism is simulated on NS2-Mobiwan2 extensions. We build the simulation model as described in Fig. 24 to evaluate the NEMO handoff latency in cases with and without Distributed Home Agents. In this model, the corresponding agent floods UDP streaming to mobile router. As shown in Fig. 25, the upper line shows that all the MNNs communicate with CN through the top HA (which near the HALR) in cellular data switch network that perform the NEMO Basic Support. The results clearly reveal that the NEMO Basic support without route optimization mechanism has the maximum total bandwidth consumption. The NEMO with proposed MRDHAD mechanism routed less flows in the whole network. Through the simulation result, the performance of our NEMO-DHA Approx-RO is as good as the ideal RO. In
NS2 simulation, the RR procedure delay (RRD) is neglected in ideal RO because it performs the binding update procedure without doing routing routability. In simulation scenario, we could conclude the results as shown in Table 2. The simulation results of system throughput is depicted in Fig. 26. The average loading is shown in Fig. 27. System delay variance is shown in Fig. 28. The NEMO-DHA has less delay time because MR always attach on home agent which is near to its domain. Also, jitter is used to analyze the packet delay variance. As shown in Table 3, the scenario of RO and MRDHAD generates more packet loss than the scenario of NO_RO because of doing extra binding procedures. But these packet losses could be recovered from preserved cache space in HA. In this case, MR-DHAD scenario Table 2 Impact of deploying multiple HAs
Fig. 23 NEMO-DHA CA inner MR case
Fig. 24 Simulation model
Router
HA
Router
Router 3
Single HA in
Best
3
3
3
network
Normal
5
3
7
7
Worst
5
3
7
7
Router
DHA
Router
DHA 3
Two HAs in
Best
3
3
3
network
Normal
5
3
5
3
Worst
5
3
5
3
Routing optimization over network mobility with distributed home agents as the cross layer consideration Fig. 25 Throughput
Fig. 26 System throughput
Table 3 Packet loss in simulation scenario Packet Field ID
No_RO
MR-DHAD
RO
Packets sent
3238
3197
3158
Packet droped
493
596
663
Packet loss rate
15.23%
18.64%
20.99%
The simulation results show that NEMO-DHA solution could provides approximate RO performance in many aspects in the small scale scenario. In this case, the simulation results comply with the analytical model.
6 Conclusion still has 2.35% improvement which is superior to RO scenario.
As the new elements in network, the MRs have the characteristic of integrating and adding more features to act as
C.-S. Li et al. Fig. 27 System average loading
Fig. 28 System delay variance
Distributed Home Agents; however, NEMO RO problem is a challenging problem. We provide the alternative approach to decrease MN-HA, CN-HA number of route hops, provide less delay time and packet loss without the need of RO procedure, economize the bandwidth and avoid most of RO problems. In addition, we suggest the Cross-Layer issue of NEMO by using R-bit should be considered since NEMO wireless connection acts a backbone. Finally, it can reach the goal of making less handoff latency. Based on our proposal, any new wireless technology used as the co-
operating partner can benefit easily from approximate-RO with the corresponding network. The profitable business model may increase the adoption of next generation Internet protocol for cellular vendors. Our proposal adopts the NEMO anycast for distributed Home Agents, which means the MR with the capability of finding the nearest HA. It could be extended to other tunneling-based RO for finding nearest CR to support NEMO RO or relaying Reverse Routing Header (RRH) [19] support request to RRH-supported HA.
Routing optimization over network mobility with distributed home agents as the cross layer consideration
References
1. Johnson, D., Perkins, C., Arkko, J. (2004). Mobility support in IPv6. RFC 3775, June 2004. 2. Devarapalli, V., Wakikawa, R., Petrescu, A., & Thubert, P. (2005). NEMO basic support protocol. RFC 3963, Jan 2005. 3. Nuorvala, V., Petander, H., & Tuominen, A. (2005). MIPL mobile IPv6 for Linux, Feb 2005. http://www.mobile-ipv6.org/. 4. KAME, SHISA, Dec. 2004. http://www.mobileip.jp/. 5. Nautilus6. www.nautilus6.org. 6. Manner, J., & Kojo, M. (2004). Mobility related terminology. RFC 3753, June 2004. 7. Futuresoft protocol stacks. www.futsoft.com. 8. Ng, C., Thubert, P., & Ohnishi, H. (2005). Taxonomy of route optimization models in the NEMO Context. Draft 04, Feb 2005. 9. Zhao, F., Wu, S.F., Davis, U.C., Jung, S. (2005). NEMO route optimization problem statement and analysis. Draft 1, Feb 2005. 10. Watari, M., & Ernst, T. (2004). Route optimization with nested correspondent nodes. Draft 00, Oct 2004. 11. Clausen, T., Baccelli, E., & Wakikawa, R. (2004). NEMO route optimization problem statement. Draft 0, Oct 2004. 12. Thubert, P., Wakikawa, R., & Devarapalli, V. (2005). Global HA to HA protocol. Draft 01, Oct 2005. 13. Paik, E. K., Cho, H.-S., Ernst, T., & Choi, Y. (2004). Load sharing and session preservation with multiple mobile routers for large scale network mobility. Advanced information networking and applications. 14. Esaki, H. (2004). Multi-homing and multi-path architecture using mobile IP and NEMO framework. Symposium on Applications and the Internet. 15. Yen, Y.-S., Hsu, C.-C., & Chao, H.-C. (2004). Dynamic home agent discovery by global network anycast on mobile IPv6. TANET. 16. Hinden, R., & Deering, S. (2006). Internet protocol version 6 addressing architecture. RFC4291, Feb 2006. 17. Ng, C., Paik, E., & Ernst, T. (2007). Analysis of multihoming in network mobility support. Draft 07, Feb 2007. 18. Galli, S., Morera, R., & McAuley, A. (2004). An analytical approach to the performance evaluation of mobility protocols: the handoff delay case. In IEEE vehicular technology conference, VTC’04, May 2004. 19. Thubert, P., & Moteni, M. (2007). IPv6 reverse routing header and its application to mobile networks. Draft 07, Feb 2007. 20. Ng, C., & Hirano, J. (2004). Securing nested tunnels optimization with access router option. Draft 01, Jul 2004. 21. Website (2009). http://www.ti-wmc.nl/. 22. Shakkottai, S., Rappaport, T. S., & Karlsson, P. C. (2003). Crosslayer design for wireless networks. IEEE Communications Magazine, Oct. 2003. 23. Tseng, C.-C. et al. (2005). Topology-aided cross-layer fast handoff designs for IEEE 802.11/mobile IP environments. IEEE Communication Magazine, 43(12), 156–163. 24. Zhu, F., & McNair, J. (2005). Cross layer design for mobile IP handoff. In IEEE VTC ’05-Spring (vol. 4, pp. 2255–2259). May– June 2005. 25. Han, Y.-H., Jang, H., Choi, J.-H., Park, B., & McNair, J. (2007). A cross-layering design for IPv6 fast handover support in an IEEE 802.16e wireless MAN. IEEE Network, November/December 2007.
Chung-Sheng Li graduated in June, 1994 from the Institute of Information Engineering National ChengKung University, Taiwan. He is currently a Ph.D. student of Department of Electrical Engineering National Dong-Hwa University. His research interest includes IPv6, Wireless Network, Mobile Computing and Network Modeling and Simulation, etc.
Fred Lin received his M.S. degree from the Department of Electronic Engineering, National Dong Hwa University in 2005, with interests in P2P and Network Mobility. He is currently employed by Delta Network, where he could make dailyused wire&wireless network equipments.
Han-Chieh Chao is a jointly appointed Full Professor of the Department of Electronic Engineering and Institute of Computer Science & Information Engineering, National Ilan University (NIU), ILan, Taiwan. He also serves as the Dean of the College of Electrical Engineering & Computer Science for NIU and Director of Computer Center for Ministry of Education. Currently he holds the joint professorship of the Department of Electrical Engineering, National Dong Hwa University, Hualien, Taiwan and the honorary adjunct professorship of the Beijing Jiaotong University, China. His research interests include High Speed Networks, Wireless Networks, IPv6 based Networks, Digital Creative Arts and Digital Divide. He received his M.S. and Ph.D. degrees in Electrical Engineering from Purdue University in 1989 and 1993 respectively. He has authored or co-authored 4 books and has published about 140 refereed professional research papers. He has completed 50 M.S.E.E. and 1 Ph.D. thesis students. Dr. Chao has received many research awards, including Purdue University SRC awards, and NSC research awards (National Science Council of Taiwan). He also received many funded research grants from NSC, Ministry of Education (MOE), RDEC, Industrial Technology of Research Institute, Institute of Information Industry and FarEasTone Telecommunications Lab. Dr. Chao has been invited frequently to give talks at national and international conferences and research organizations. Dr. Chao is also serving as an IPv6 Steering Committee member and co-chair of R&D division of the NICI (National Information and Communication Initiative, a ministry level government agency which aims to integrate domestic IT and Telecom projects of Taiwan), Co-chair of the Technical Area for IPv6 Forum Taiwan, the executive editor of the Journal of Inter-
C.-S. Li et al. net Technology and the Editor-in-Chief for International Journal of Internet Protocol Technology and International Journal of Ad Hoc and Ubiquitous Computing. Dr. Chao has served as the guest editors for Mobile Networking and Applications (ACM MONET), IEEE JSAC, IEEE Communications Magazine, Computer Communications, IEE Proceedings Communications, Telecommunication Systems, Wireless
Personal Communications, Computer Journal and Wireless Communications & Mobile Computing. Dr. Chao is an IEEE senior member, a Fellow of the Institution of Engineering and Technology (FIET), and Chartered Fellow of British Computer Society (FBCS). Home page: http://www.ndhu.edu.tw/~comput/HCC/index.htm.
