Cost-efficient network mobility scheme over proxy mobile IPv6 network

www.ietdl.org Published in IET Communications Received on 1st February 2011 Revised on 19th July 2011 doi: 10.1049/iet-com.2011.0109

Special Section: Green Technologies for Wireless Communications and Mobile Computing ISSN 1751-8628

Cost-efficient network mobility scheme over proxy mobile IPv6 network S. Jeon Y. Kim Electronic Engineering, Soongsil University, Seoul, South Korea E-mail: [email protected]

Abstract: As Wi-Fi handheld devices providing a variety of data and multimedia services become more widespread, and as the global demand for internet access everywhere rapidly increases, energy-efficient network mobility (NEMO) technology is attracting great attention with respect to the latest research in mobile networking. The previous research relating to energyefficient networks in mobile networking has focused on routing protocol or topology control. In NEMO, cost-efficient protocol design reduces unnecessary costs because of mobility signalling and packet delivery, and saves limited network bandwidth in both wired and especially wireless, which is one of the important issues of energy efficiency. In this study, the authors propose a cost-efficient proxy router-based NEMO scheme (PR-NEMO) providing efficient mobility management and packet delivery method. In PR-NEMO, a mobile network is formed with a unique ID and the information of mobile network nodes (MNNs) within the mobile network anchored at local mobility anchor (LMA) are distributed to attached mobile access gateway (MAG). The authors analyse the performance of the PR-NEMO with the N-PMIPv6 and the rNEMO in terms of total cost; that is, sum of location update cost and packet-tunnelling cost. Numerical results demonstrate that the PR-NEMO is the most cost-efficient scheme, under various conditions, among the three schemes evaluated.

1

Introduction

Network mobility (NEMO) is a novel concept for handling a group of nodes within a moving vehicular area; in particular, it provides an effective way for wireless devices to access the internet through an intermediate router connecting to an external wireless wide access network as shown in Fig. 1. The recent rapid propagation of handheld Wi-Fi devices and the increasing demand for Internet access everywhere have made NEMO technology much more noticeable [1]. To execute NEMO deployment, the NEMO basic support protocol (NEMO-BSP) [2] was created by the internet engineering task force (IETF) by installing an extended mobile IPv6 (MIPv6) host protocol [3] on a mobile router (MR), which is in charge of mobility management for mobile network nodes (MNNs). The NEMO-BSP deploys a home agent (HA) for the MR and then all packets addressed to an MNN are delivered via the MR’s HA, leading to packet delivery overhead. Even if the MNN is visiting a mobile node that is moved from another MIPv6 network, the packet overhead is more severe and a mobile terminal is involved to update its new location to its HA. These drawbacks are linked to the issue of cost-efficiency when considering a frequent handoff and a large number of MNNs within a mobile network. A few of energy-efficient network research in mobile networking have been studied on routing protocol [4], topology control [5], power consumption of wireless interface [6]. However, in NEMO, cost-efficient protocol design that reduces the unnecessary network bandwidth because of mobility signalling and 2656 & The Institution of Engineering and Technology 2011

packet delivery is one of the important energy-efficient issues [7, 8]. In recent years, a few studies on applicable NEMO over Proxy Mobile IPv6 (PMIPv6) [9] have been attempted, because the PMIPv6 provides a network-based enhanced local mobility and does not require host-stack involvement: the NEMO-enabled PMIPv6 (N-PMIPv6) and relay-based NEMO (rNEMO). The N-PMIPv6 [10] introduces the moving mobile access gateway (mMAG) instead of an MR defined in the NEMOBSP. The mMAG is also in charge of mobility management for MNNs, so that the MAG does not need to keep track of the MNNs. This causes a local mobility anchor (LMA) to perform a recursive lookup, and then a packet addressed to an MNN is sent through a nested tunnel as shown in Fig. 2. It changes the normal operation of the LMA and leads to serious packet-tunnelling cost. In the rNEMO [11], a simple relay station – which can be either an amplify-and-forward (AF) or a decode-and-forward (DF) relay station – is employed to mitigate the packettunnelling costs incurred by simply applying the NEMOBSP to the PMIPv6 network. As the relay station has no functionality for mobility management, a MAG is required to perform location updates for individual MNNs whenever a mobile network moves to the cell region of another MAG, as shown in Fig. 3. Consequently, both approaches have trouble in facilitating cost-efficient NEMO operation. Therefore we propose a cost-efficient NEMO protocol, called the proxy router-based NEMO (PR-NEMO) in the PMIPv6 network. The proxy router (PR) is employed to IET Commun., 2011, Vol. 5, Iss. 18, pp. 2656–2661 doi: 10.1049/iet-com.2011.0109

www.ietdl.org changes the MAG IP address of all the nodes within the mobile network to the new MAG IP address to which the PR attaches to enable an efficient handoff management with no additional signalling. After the LMA processes handoff signalling for the PR, the LMA announces the stored MNNs’ information to a new MAG, to which the PR attaches. This paper focuses on how to jointly reduce protocol overhead required for location update and packet delivery in order to achieve cost-efficient NEMO solution. For the performance evaluation, we define the total cost, which is the sum of the location update cost and the packettunnelling cost. The rest of the paper is organised as follows: Section 2 describes the protocol operation of the proposed PRNEMO. In Section 3, we analyse the performance of the PR-NEMO in comparison with that of the N-PMIPv6 and the rNEMO. Section 4 presents the numerical results. Finally, we conclude with Section 5. Fig. 1 Concept of NEMO

2 2.1

Fig. 2 Signalling and packet delivery procedure of the N-PMIPv6

Proposed scheme Attachment operation in the PR-NEMO

Fig. 4 shows the attachment procedure of the PR and the MNN in the proposed PR-NEMO. When a PR approaches the PMIPv6 network, this event is detected by the MAG, which then sends a proxy binding update (PBU) message containing ‘B’ and ‘N’ flags set to ‘1’ towards the LMA. The ‘B’ flag, which requests the allocation of a group ID (GID), is derived from the bulk re-registration mechanism to reduce the excessive signalling cost required to extend the binding lifetime of individual mobile hosts [12]. However, we employ this concept to support efficient handoff in a mobile network. Thus, we added an ‘N’ flag, which requests the allocation of a new GID to the reserved field within the PBU message and proxy binding acknowledgment (PBA) message. To support the given procedure, a MAG should know that the attached node is the router or a mobile host. Corresponding information can be obtained from the security association between the MAG and the AAA.

Fig. 3 Signalling and packet delivery procedure of the rNEMO

perform proxy forwarding of data packets between an MNN and a MAG and node information to a MAG on behalf of MNN; the LMA manages a PR and MNNs within one mobile network as a unique proxy router group ID (PRGID). When the PR hands off to new MAG, the LMA IET Commun., 2011, Vol. 5, Iss. 18, pp. 2656–2661 doi: 10.1049/iet-com.2011.0109

Fig. 4 Attachment operation in the PR-NEMO 2657

& The Institution of Engineering and Technology 2011

www.ietdl.org When an MNN attaches to the PR, it sends its MNN-ID to the PR using a router solicitation (RS) message. (According to the RFC 5213 [8], ‘the specification can operate without linklayer indications of node attachment and detachment to the link.’ In accordance with the definition, the RS message is used to transmit the MN-ID.) To deliver the RS message to the MAG, the PR has forwarding functionality for the signal and data packets. The MNN-ID can be used as a MAC address or a network address identifier (NAI) [13]. The NAI is more suitable for PR-NEMO because the PR does not need to handle the received RS to forward it to the MAG. To distinguish the origin of the forwarding RS message at the MAG, a ‘P’ flag is added in this paper. On receiving the RS message, the MAG learns that the attached MNN belongs to the PR by confirming that the ‘P’ flag is set to ‘1’. The RS message contains the MNN’s ID. To manage the attached node under the PR, a PBU message sent by the MAG should include the MNN-ID, the PR-ID of the attached PR and a ‘B’ flag set to ‘1’. The LMA allocates the home network prefix (HNP) for the MNN and makes a group of the MNN and the PR with the PR-GID. After receiving the PBA message, the MAG adds the MNN to the PR-GID group. The allocated HNP is delivered to the PR through a router advertisement (RA) message with the ‘P’ flag set to ‘1’. Then the RA message with the ‘P’ flag set to ‘0’ is finally delivered to the MNN. Fig. 5 shows an extended binding cache entry (BCE) within the LMA for the PR-NEMO mechanism. For example, three nodes (Node-1, Node-2 and Node-3) are registered at the same LMA. Node-1 is a PR with PR-GID1, whereas Node-2 is an MNN belonging to the same mobile network. However Node-3 does not belong to any mobile network but is directly attached to the PMIPv6 network. When the LMA receives the data packet from an external router, it checks the BCE and then forwards them to the appropriate MAG address using the PMIPv6 tunnel. 2.2

Handoff operation in the PR-NEMO

Fig. 6 Handoff operation in the PR-NEMO

PR-GID with the PR-ID. If the PR-ID and PR-GID of the PR exist on the LMA BCE, the LMA changes the MAG IP address fields of all the nodes that have same PR-GID to the N-MAG’s IP address. The LMA then sends the PBA message, including the MNN-IDs and MNN-HNPs, as well as the PR-HNP and PR-GID, to the N-MAG, because the N-MAG has no information about the MNNs that belong to the PR. Information delivery makes it possible in an MAG to forward the data packets to the PR with no tunnelling procedure. The N-MAG transmits the RA message with the HNPs of all MNNs to the PR, followed by a normal PMIPv6 operation. This behaviour can make the PBA message slightly larger, depending on the number of MNNs within a vehicle. However, when we consider that one MNN has one or more sessions and one session has dozens of packets at minimum, the influence because of increased PBA message sizes is not greater than the packet tunnelling cost generated from one MNN. Detailed results are shown in the numerical results in Section 4.

Fig. 6 shows the handoff operation for a mobile network when a PR moves to the next MAG (N-MAG) from the previous MAG (P-MAG). To support efficient handoff, only the PR is involved with the handoff operation. When the PR is moving to the N-MAG, the P-MAG detects the PR’s detachment from layer-2 signals, then it initiates a de-registration process for the PR. The N-MAG that detects a PR’s attachment performs the location update procedure in the same way as the PR’s initial attachment operation. From AAA procedure, the N-MAG learns that the attached node is a proxy router for NEMO service; then, the N-MAG sends the PBU signal message with ‘B’ and ‘N’ flags to the LMA. On receiving the PBU signal, the LMA looks up the corresponding

2.3 Node handoff operation from the PR to the outer PMIPv6 network

Fig. 5 Extended BCE within the LMA

Fig. 7 Node handoff operation in the PR-NEMO

2658 & The Institution of Engineering and Technology 2011

For node mobility operation, the PR is required to inform the MAG of the MNN’s detachment received from the MNN. As

IET Commun., 2011, Vol. 5, Iss. 18, pp. 2656–2661 doi: 10.1049/iet-com.2011.0109

www.ietdl.org shown in Fig. 7, when an MNN moves out from a mobile network, the PR detects the MNN’s detachment from a layer-2 signal and then sends an RS message to the current MAG. This RS message includes ‘P’ and ‘D’ flags set to ‘1’ and an MNN-ID. The ‘D’ flag is defined in this document to announce the MNN’s movement to the MAG. When the MAG receives the RS message with the ‘P’ and ‘D’ flags set to ‘1’, it determines that the corresponding MNN-ID, in a received RS message, moved out from the PR and it sends a PBU message for de-registration to the LMA. The LMA performs the de-registration and removes the MNN from the group to which the PR belongs; then it deletes the stored GID for the MNN and sends a PBA message to the MAG. When the MAG receives the PBA message from the LMA, it also removes the MNN from the group to which the PR belongs and the MNN-related entry information in the proxy binding update list (PBUL). When the MNN approaches the MAG, the MAG detects the node’s movement and performs the node attachment procedure followed as a normal PMIPv6 operation.

3

N-PMIPv6 = mC T (k RSmMAG−MAG + k RAMAG−mMAG CLU + t AAAMAG−AAA + 2t LUMAG−LMA )

PR-NEMO = mC T (k RSPR−MAG + k RAMAG−PR CLU

+ t AAAMAG−AAA + 2t LUMAG−LMA + dtdLMA−MAG (LMNN−ID + LMNN−HNP )) 3.2

The packet-tunnelling cost denotes the size of the IP tunnel header of the data packets delivered from the correspondent node (CN) to the MNN. According to Little’s law [15], the number of packets delivered during T is approximated by lSTE(S). Therefore the CPT of the rNEMO is derived by rNEMO = lS TE(S)(tdLMA−MAG LT ) CPT

N−PMIPv6 = lS TE(S)(tdLMA−MAG 2LT CPT

+ kdMAG−mMAG LT )

where d is the number of MNNs and t and k denote the unit transmission costs in a wired and a wireless link, respectively. In general, since the transmission cost in a wireless link is IET Commun., 2011, Vol. 5, Iss. 18, pp. 2656–2661 doi: 10.1049/iet-com.2011.0109

(5)

The CPT of the PR-NEMO is the same as that of the rNEMO because no tunnel header is used except for the PMIPv6 tunnel header between the MAG and the LMA

4

(6)

Numerical results

To compute the total cost of the three schemes, some parameter values are derived in the literature [16] as shown in Table 1. LT is 40 bytes and LP is assumed to be 1500 bytes. The signalling sizes for RS/RA/AAA and PBU/PBA are 70/110/100 bytes and 112/96 bytes, as specified in [9]. Table 1

(1)

(4)

In the N-PMIPv6, when an LMA receives the packets addressed to the MNN, it adds two tunnel headers to the received packets, which then are sent to the MAG. The MAG removes only the outer header and forwards the rest to the mMAG. Therefore we can express the CPT by

PR−NEMO CPT = lS TE(S)(tdLMA−MAG LT )

In the rNEMO scheme, when a mobile network moves to another MAG, the MAG performs a location update operation with the LMA by sending/receiving a PBU/PBA message for individual MNNs within the PR. An MNN exchanges the RS/RA message with the PR as the number of MNNs. The average number of subnet crossings during T is equal to mCT, as defined in [14]. Therefore the location update for rNEMO can be represented by

(3)

Packet-tunnelling cost

Location update cost

rNEMO CLU = mC T d(k RSMNN−MAG + k RAMAG−MNN + t AAAMAG−AAA + 2t LUMAG−LMA )

(2)

In the proposed PR-NEMO scheme, the PR also performs a location update operation on behalf of the MNNs. However the MNN-IDs and MNN-HNPs stored at the LMA BCE should be delivered to the newly attached MAG by the PBA message. To add the length of the MNN-ID and MNN-HNP options as the number of MNNs, we separately add dtdLMA-MAG.(LMNN-ID + LMNN-HNP) to the CLU of the PR-NEMO. Therefore the CLU can be calculated by

Performance analysis

In this section, we analyse the performance of three schemes. In the literature [10], they consider throughput performance in comparison with hybrid environment such as NEMOBSP + MIPv6 + PMIPv6. However, we handle three schemes in same protocol domain, and our main concern is to reveal how much we can reduce the costs generated from network side for supporting each NEMO scheme. Thus, for evaluating the cost, we analyse the total cost (CTOT), which is the sum of the location update (LU) cost (CLU) and the packet-tunnelling (PT) cost (CPT). Each cost is defined by the product of the message size and the hop distance from a bandwidth perspective. Under this definition, the router processing cost is not considered. For the analytical model, we make assumptions as follows. The inter-session arrival time follows an exponential distribution with a rate of lS , and the average session length, in packets, is E(S). The MAG subnet crossing rate follows a general distribution with a mean rate of mC , as presented in [14]. da2b denotes the hop distance between network entities a and b. RSa2b and RAa2b denote the unit cost of the RS and RA messages between a and b as da2b.(LRS) and da2b.(LRA), where Lg denotes the message sizes of g, respectively. LUa2b denotes the unit cost of the PBU/PBA message between a and b as da2b.(LPBU + LPBA). 3.1

larger than that in a wired link, k is larger than t. In the N-PMIPv6 scheme, only the mMAG performs a location update operation on behalf of the MNNs. CLU is constant regardless of the number of MNNs within the PR. Therefore the corresponding CLU can be calculated by

Parameters for numerical results

Parameter dMNN-PR(mMAG) dPR(mMAG)-MAG dMAG-AAA

Value

Parameter

Value

1 1 1

dMAG-LMA d/k/t T, s

4 10/2/0.5 1000

2659


www.ietdl.org

Fig. 8 Total cost for the session arrival rate and the MAG subnet crossing rate a No. of MNN ¼ 30 b E(s) ¼ 20

The option message sizes for MNN-HNP and MNN-ID are 20 and 32 bytes, respectively [8, 13]. Fig. 8a shows the total cost for different session arrival rates. An increased session arrival rate affects the packet delivery cost; therefore, the N-PMIPv6 is strongly influenced by a rising session arrival rate. The total cost of the rNEMO is greater than that of the PR-NEMO, even though the rNEMO has the same packet-tunnelling cost as the PR-NEMO. This is because the rNEMO performs an individual location update process in relation to the number of MNNs. From Fig. 8b, we confirm that the rNEMO is highly vulnerable to frequent IP handoff events. The total cost of N-PMIPv6 and the PR-NEMO also increases, but the increases result from the PR’s handoff signalling. Fig. 9 shows the total cost to check the influences as the number of MNNs increases. As the number of MNNs increases, the total cost of the N-PMIPv6 rapidly increases compared to that of the other two schemes, because of increasing packet-tunnelling costs. In the PR-NEMO, the PBA message size increases for the number of MNNs, because it contains more data as the number of MNNs grows, but the packet-tunnelling cost against increasing

Fig. 9 Total cost as a function of the number of MNNs

Fig. 10 Total cost as a function of SMR a E(s) ¼ 20 b No. of MNN ¼ 40 2660 & The Institution of Engineering and Technology 2011


www.ietdl.org MNNs also increases at the same time. If E(S) is assumed to be at least 10 packets per MNN, the results show that increased PBA sizes do not affect the total cost much. When we compare the total cost of the PR-NEMO and the rNEMO, the results also show that the increased PBA sizes in PR-NEMO are lower than the increasing location update costs. To show the total cost in a diverse network environment, we use the session-to-mobility ratio (SMR), defined as the ratio of the session arrival rate to the handoff rate. If the SMR becomes greater, it means that the session activity is relatively higher than the handoff rate. The total cost of rNEMO is greatly influenced by the mobility rate, while the N-PMIPv6 does not make a big difference, depending on the SMR value. In the PR-NEMO, the total cost increases slightly in accordance with the SMR increases in the case of Fig. 10b.

5

Conclusions

This paper proposed a cost-efficient NEMO protocol called the PR-NEMO over PMIPv6 network. The PR was introduced to perform proxy forwarding of data packets between mobile network and PMIPv6 and notify the existences of MNNs. The LMA groups all the MNNs that belong to same mobile network with a unique ID. When the PR attaches to a new MAG, the cache of MNNs belonging to the PR is announced to the MAG from the LMA. PR-NEMO provides efficient mobility management and packet delivery method. As a result, PR-NEMO reduced both location update overhead and packet-tunnelling overhead. By analysing and evaluating the total cost of the PR-NEMO with N-PMIPv6 and rNEMO, we confirmed that the PR-NEMO is a highly cost-efficient NEMO solution compared to other schemes under various conditions. It is also conjectured that the PRNEMO is a suitable solution for moving Wi-Fi hotspot environments.

6

Acknowledgments

This research was partially supported by the KCC (Korea Communications Commission), Korea, under the R&D program supervised by the KCA (Korea Communications Agency) (KCA-2011-08913-05001) and by the MKE (The


Ministry of Knowledge Economy), Korea, under the Convergence-ITRC (Convergence Information Technology Research Center) support program (NIPA-2011 C61501101-0004) supervised by the NIPA (National IT Industry Promotion Agency).

7

References

1 Pack, S., Shen, X., Park, J.W., Pan, J.: ‘Mobility management in mobile hotspots with heterogeneous multihop wireless links’, IEEE Commun. Mag., 2007, 45, (9), pp. 106– 112 2 Devarapalli, V., Wakikawa, R., Petrescu, A., Thubert, P.: ‘Network mobility (NEMO) basic support protocol’. IETF RFC 3963, January 2005 3 Johnson, D., Perkins, C., Arkko, J.: ‘Mobility support in IPv6’. IETF RFC 3775, June 2004 4 Zabin, F., Misra, S., Woungang, I., Rashvand, H.F., Ma, N.-W., Ali, M.A.: ‘REEP: data-centric, energy-efficient and reliable routing protocol for wireless sensor networks’, IET Commun., 2007, 2, (8), pp. 995–1008 5 Venkataraman, G., Emmanuel, S., Thambipillai, S.: ‘Energy-efficient cluster-based scheme for failure management in sensor networks’, IET Commun., 2007, 2, (4), pp. 528– 537 6 Namboodiri, V., Gao, L.: ‘Energy-efficient VoIP over wireless LANs’, IEEE Trans. Mob. Comput., 2010, 9, (4), pp. 566– 581 7 Lee, J., Ernst, T., Chung, T.: ‘Cost analysis of IP mobility management protocols for consumer mobile devices’, IEEE Trans. Consum. Electron., 2010, 56, (2), pp. 1010–1017 8 Makaya, C., Pierre, S.: ‘An analytical framework for performance evaluation of IPv6-based mobility management protocols’, IEEE Trans. Wirel. Commun., 2008, 7, (3), pp. 972 –983 9 Gundavelli, S., Leung, K., Devarapalli, V., Chowdhury, K., Patil, B.: ‘Proxy mobile IPv6’. IETF RFC 5213, August 2008 10 Soto, I., Bernardos, C.J., Calderon, M., Banchs, A., Azcorra, A.: ‘NEMO-enabled localized mobility support for internet access in automotive scenarios’, IEEE Commun. Mag., 2009, 47, (5), pp. 152–159 11 Pack, S.: ‘Relay-based network mobility support in proxy mobile IPv6 networks’. Proc. IEEE CCNC 2008, January 2008 12 Abinader, F., Gundavelli, S., Leung, K., Krishnan, S., Premec, D.: ‘Bulk re-registration for proxy mobile IPv6’, draft-ietf-netext-bulk-reregistration-01.txt, July 2010 13 Aboba, B., Beadles, M., Arkko, J., Eronen, P.: ‘The network access identifier’. RFC 4282, December 2005 14 Pack, S., Shen, X., Mark, J.W., Pan, J.: ‘Adaptive route optimization in hierarchical mobile IPv6 networks’, IEEE Trans Mob. Comput., 2007, 6, (8), pp. 903– 914 15 Kleinrock, L.: ‘Queueing systems volume 1: theory’ (John Wiley & Sons, 1975) 16 Jeon, S., Kang, N., Kim, Y.: ‘Enhanced predictive handover for fast proxy mobile IPv6’, IEICE Trans Commun., 2009, E92-B, (11), pp. 3504– 3507

2661
