Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks

Information Systems ] (]]]]) ]]]–]]]

Contents lists available at ScienceDirect

Information Systems journal homepage: www.elsevier.com/locate/infosys

Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks Chulhee Cho a, Jeong-Jin Kang b, Jongpil Jeong c,n a b c

Graduate School of Information and Communications, Sungkyunkwan University, Seoul 110-745, Republic of Korea Department of Information and Communication, Dong Seoul University, Seongnam, Gyeonggi-Province 461-714, Republic of Korea College of Information and Communications, Sungkyunkwan University, Suwon, Kyunggi-do 440-745, Republic of Korea

a r t i c l e i n f o

Keywords: PMIPv6 Paging Handoff DNS Global Mobility Management

abstract Proxy Mobile IPv6 (PMIPv6) is designed to provide a network-based localized mobility management protocol, but it does not handle the global mobility of hosts. In this paper, we propose a location management scheme based on Domain Name System (DNS) for PMIPv6. In this proposed scheme, DNS as a location manager provides PMIPv6 for global mobility. In addition, a paging extension scheme is introduced to PMIPv6 in order to support large numbers of mobile terminals and enhance network scalability. To evaluate the proposed location management scheme, we establish an analytical model, also formulate the location update and the paging cost, and analyse the influence of the different factors on the total signalling cost. The performance results show how the total signal cost changes under various parameters. & 2014 Elsevier Ltd. All rights reserved.

1. Introduction The network-based localized mobility management (NETLMM) working group standardized a NETLMM protocol called PMIPv6 [1], where local IP mobility is handled without involvement from the mobile node. Compared to host-based mobility management approaches such as Mobile IPv6 (MIPv6), PMIPv6 has fundamental advantages [2]. The MN is not required to participate in any mobilityrelated signalling, which reduces signalling cost in wireless links and takes advantage of wireless resources efficiently. Secondly, PMIPv6 does not require any modification of the MN, because it could distribute services to various types of terminals. Finally, as a network-based mobility management protocol, PMIPv6 makes it possible to handle the

n

Corresponding author. E-mail addresses: [email protected] (C. Cho), [email protected] (J.-J. Kang), [email protected] (J. Jeong).

network easily, because it can control the network traffic [3]. However, PMIPv6 has some disadvantages as well. One is that PMIPv6 does not currently provide the global mobility of hosts [1,4]. PMIPv6 is designed to give network-based mobility management support to an MN in a topologically localized domain. The current PMIPv6 does not make an appropriate solution for the case of inter-mobility from one domain to others. Thus, we have to introduce a network entity to be in charge of a global location manager to PMIPv6. Most methods of communication in the current Internet start with a name lookup via DNS in order to translate a domain name into an IP address. DNS is used widely, so it could globally locate an MN. [5] suggested a feasible solution without new entities using the DNS as a location manager. Thus, in order to support global mobility in PMIPv6, we bring in a DNS that performs the role of a location manager. Another drawback is that PMIPv6 supports registration but not paging [6], which could reduce the location update cost and power consumption of MNs. Paging technology has other features

http://dx.doi.org/10.1016/j.is.2014.08.001 0306-4379/& 2014 Elsevier Ltd. All rights reserved.

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

2

that expand the scalability of the protocol and handle numerous mobile nodes used widely in cellular systems [7,8]. Therefore, we propose a paging extension scheme in order to optimize mobility management performance in PMIPv6. In this paper, we propose a location management scheme for PMIPv6 called D-PMIPv6, which is based on DNS and a paging extension scheme. In addition, in order to evaluate the location management scheme, we establish an analytical model. From this model, we formulate the location update cost and the paging cost and explain how the PMIPv6 domain, mobility rate, session arriving rate, and active mode rate affect the total signalling cost with various parameters. We also propose a DNS security extension called DNSSEC in order to improve the security. The DNS protocol was designed without security as a central concern, and a variety of possible attacks against DNS have been identified [9]. The rest of this paper is organized as follows. Section 2 shows a brief overview of PMIPv6. Section 3 gives the details for location management mechanism based on DNS. Section 4 discusses the security analysis, while Section 5 formulates an analytical model and the signalling cost. Section 6 evaluates the performance results. Finally, in Section 7 we present the conclusion.

2. Overview of PMIPv6 In this section, we give a short overview of PMIPv6, as shown in Fig. 1. PMIPv6 is designed to provide a networkbased IP mobility PMIPv6 domain architecture. The PMIPv6 domain consists of two core functional elements: the Local Mobility Anchor (LMA) and the Mobile Access Gateway (MAG). The LMA manages the Home Agent and Binding Cache Entry of a currently registered MN. Meanwhile, the LMA handles possibility. Wherever the MN moves within the PMPv6 domain, it is managed by the LMA. The main role of the MAG is to detect the MN's movements and manage the mobility-related signalling on behalf of the MN. In PMIPv6, as an MN first reaches a MAG in the PMIPv6

domain, it sends a router solicitation (RS) message to the MAG. After the MN passes an access authentication procedure, the MAG obtains the MN's profile, which includes the MN identifier, which is the LMA's address. Afterwards, the MAG sends a proxy binding update (PBU) message on behalf of the MN to its LMA. On receiving the PBU message, the LMA sends the proxy binding acknowledgement (PBA) message including the MN's home network prefix to the MAG. At the same time, the LMA records the MN's information in the binding cache entry and establishes a bidirectional tunnel between the LMA and the MAG. Once the MAG receives the PBA message, it sends a router advertisement (RA) message to the MN and sets up a bidirectional tunnel. After receiving the RA message, which contains the home network prefix, the MN can configure its proxy home address (pHoA). After an MN obtains a pHoA in the PMIPv6 domain, it can send and receive data traffic with its pHoA. The LMA receives all of the data packets sent by the MN to the MN, and then forwards the received packets to the MAG through the tunnel. After receiving the packets, the MAG on the other end of the tunnel removes the outer header and forwards the packets to the MN. 3. A novel location management scheme based on DNS in PMIPv6 In this section, we suggest a location management scheme based on DNS for PMIPv6. Under this new scheme, DNS is in charge of the global location manager, and an extended location update scheme needs to replace the currently existing one. In addition, a paging scheme is introduced to PMIPv6. 3.1. DNS as location manager for PMIPv6 PMIPv6 has a feature that does not support global mobility of hosts [10]. Under this circumstance, we need to consider bringing in a network entity as the global location manager. DNS is already a part of the existing Internet infrastructure, and most communications in the traditional Internet start with a name lookup via DNS to translate the Fully Qualified Domain Name (FQDN) to an IP address. Therefore, DNS could be supposed as a location manager for supporting global mobility in PMIPv6. The DNS as location manager contains two operations: location update and location search [6]. 3.1.1. Location update In PMIPv6, when an MN hands off in the same PMIPv6 domain, there is no need to change its address. As a result, the FQDN-to-IP entry in the DNS does not need to change either. However, if an MN moves into a different PMIPv6 domain, it obtains a new address from the current home network prefix. Then, the MN updates DNS's FQDN-to-IP entry with its new address by sending DNS UPDATE messages.

Fig. 1. PMIPv6 domain architecture.

3.1.2. Location search When communicating with an MN, the host checks for DNS's FQDN-to-IP entry. After the DNS responds with the current IP address of the MN, the host can initiate and

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

establish communication with the MN directly. Because the MN's address is part of the LMA's address prefix, data packets pass through the LMA. According to the binding cache entry, the LMA forwards data packets to the MAG through the tunnel and data packets arrive at the MN. 3.2. Location update scheme In D-PMIPv6, the location update scheme includes the MAG informing the LMA of the MN's location information and updating the DNS's record of AAA. In this section, we propose a location update scheme based on DNS. In addition, we discuss the mobile node's mode, the location update mechanism based on different modes, and the evaluation details of the location update scheme. 3.2.1. Mobile node's mode In D-PMIPv6, there are two modes of the MN's state: the active mode and the idle mode. The LMA and MAGs keep the MN's state consistent. So, the LMA's binding cache entry and the MAG's binding update list entry should add two options of the node's state and active timer. When sending or receiving data packets, the idle MN enters the active mode. When the serving MAG and LMA receive packets, they set the MN's state of the LMA's binding cache entry and the MAG's binding update list entry for the active mode. Whenever the MN has any data session, the active timer is reset. The serving MAG and LMA. When the lifetime is over, they set the mode to idle. 3.2.2. Location update scheme When an MN in active mode moves from the previously attached mobile access gateway (p-MAG) to the newly attached mobile access gateway (n-MAG) within the same PMIPv6 domain, the p-MAG sends a PBU message to the LMA in order to delete the tunnel and binding cache entry. Then, the n-MAG informs the LMA of the MN's new location information and establishes a bi-directional tunnel between the LMA. In this case, the LMA does not have to send information about the MN's movement to the DNS, because the MN's address is not changed. The mechanism is similar to Hierarchical Mobile IPv6 mobility management (HMIPv6) [11], which can reduce the signalling cost and the burden of the DNS. As an MN in active mode is handed over between PMIPv6 domains, it obtains a new address. As the MN of the active mode moving between PMIPv6 needs to receive a new address, the LMA updates the FQDN-to-IP entry of the DNS. When an MN in idle mobile moves around within the same PMIPv6 domain, the n-MAG does not need to set tunnel and create Binding Cache Entry. Furthermore, the signalling cost is reduced, because the n-MAG does not have to send a PBU message, If an MN in idle mode moves between the LMAs, the operations are the same as with an MN in active mode. 3.3. Paging scheme The paging scheme is the process of locating the MN's network. It has a feature that determines how effectively and quickly it could find out the location of the MN. In this section, we suggest a paging extension scheme for PMIPv6

3

and discuss the paging area construction, the influence of the two MN modes on the paging scheme, and the message flows of the paging scheme. 3.3.1. Paging area architecture The Internet is divided into several PMIPv6 domains, and each of them is a paging area which consists of an LMA and a few MAGs. Each paging area is controlled by the LMA and identified by a unique paging area ID, which is one of the LMA's addresses. 3.3.2. The influence of mobile node's mode on paging scheme The LMA's binding cache entry maintains the MN's state. When receiving data packets flowing into an MN, the LMA checks the MN's state. If the MN is in an active mode, the LMA forwards data packets to the MAG connecting with attaches. If the MN is in an idle mode, the LMA sends paging request messages to all the MAGs which are in the PMIPv6 domain. 3.3.3. Messages flows of paging scheme Fig. 2 illustrates the paging extension scheme based on DNS for PMIPv6. The MAG1 detects the booting of the MN, and then the MAG1 performs the registration process with the LMA. As a result, the tunnel is set up, and then the data packets are transferred to the MN. If there is no process of data packets for a long time, the MN in the active mode turns into the idle mode. Because the network traffic is concentrated on the LMA and causes a network bottleneck, the serving MAG sends idle mode notification messages to other MAGs in the paging area of the LMA1 more than the LMA. When an MN in idle mode is connected with the MAG2, the MAG2 does not set up a tunnel for the MN or send a PBU message to the LMA1. The reason why this happens is that the MAG2 knows the MN is in the idle mode. If the correspondent node (CN) wants to communicate with the MN, the CN requests the MN's address from the DNS. After acquiring the MN's address, the CN sends data packets to the LMA1 which is currently serving to the MN. The LMA knows that the MN's state is in the idle mode and sends paging request messages to all MAGs in the PMIPv6 domain. LMA1 and MAG2 create a tunnel, sending buffered data packets to the MN. The MN's location information like an MN's movement from LMA1 to LMA2 needs to be updated on LMA2 and DNS. 4. Security analysis DNS is a distributed hierarchical database that is mainly used to complete the mapping between host names and IP addresses on the Internet. Considerations of security early in the design were insufficient in the data integrity validation. With the rapid development of the Internet, DNS threats are increasingly significant, including cache poisoning attacks, denial of service attacks, and domain name hijacking attacks. In order to improve the security of the DNS, IETF introduced a data signature and PKI to the DNS, which provide data integrity and source validation and greatly improve the security of the DNS infrastructure. DNSSEC was first proposed in 1999 in RFC 2535 (Domain Name System Security Extensions) [12]. Since then, extensive research was

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

4

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

Fig. 2. Message flows of the paging scheme based on DNS for PMIPv6.

done, and relative RFCs were proposed, including 4033 (DNS Security Introduction and Requirements) [13], 4034 (Resource Records for the DNS Security Extensions) [14] and 4035 (Protocol Modifications for the DNS Security Extensions) [15]. The essence of DNSSEC is to introduce public key technology to the Domain Name System (DNS) and signing DNS responses with signatures, which will provide DNS authentication and message integrity. The principle is that before sending a DNS response, the sender uses a hash function to calculate the DNS information and obtains the “message digest”, and then encrypts the message digest with a private key. This process is called a digital signature. When the process is completed, the DNS information and its digital signature are sent together. When the DNSSEC response is received, the receiver uses the corresponding public key to decrypt the digital signature, which will produce the message digest. After that, using the same hash function with the sender to compute the message

digest again, if the two digests are the same, it can prove that the DNS information is complete and correct. DNSSEC satisfies security characteristics as follows. (1) Authenticity and integrity of the reply data: DNSSEC uses public-key cryptography (RSA, for example) to enable each zone to prove the authenticity and integrity of its DNS data. To do so, each zone creates a public-private key pair, stores the public key in a new RR type called DNSKEY RR, signs its data (in units of RRsets) using the private key, and stores the signatures at the authoritative servers in another new type of RR called RRSIG. Whenever a DNSSEC-enabled server returns an RRset, it also returns the companion signature. A resolver uses the zone's public key to verify whether a received RRset matches the signature. A match indicates that the RRset indeed originated from that zone and was not altered in transit. (2) Replay Attack Resistance: To resist replay attacks, each signature carries an expiration time specified by a definitive timestamp, and becomes invalid

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

beyond this timestamp. Accordingly, the cache discards an RRset when either its TTL or the companion signature expires, whichever comes first. (3) Authentication of denial of existence: In addition to authenticating RRsets through RRSIGs, a zone must also provide authenticated answers when it receives queries for RRsets that do not exist. Because of the desire to keep the private keys offline, upon receiving the query for a non-existing name, a zone cannot sign a denial of existence response in real time. Instead, authentication of the denial of existence is achieved in the following way. A zone first sorts all the existing names in a canonical order, and then creates an RR of a new type called NSEC for each of its names, and signs these NSEC RRs using the private key. The data portion of an NSEC RR indicates which RRsets exist under the name and identifies the next existent name in the zone. An NSEC RR together with its signature can prove the nonexistence of the queried data.

5. Analytical model for performance evaluation The performance evaluation parameter of location management is the total signalling cost, which consists of two parts: the location registration and location update signalling cost, and the paging signalling cost. In the following discussion, we develop an analytical model to evaluate the total signalling cost. This model borrows some ideas from [16]. Table 1 shows the main parameters and their descriptions.

5.1. Handover latency We dealt with handover latency and DNS update time when an MN moves within PMPv6 domains. Fig. 3 was reorganized according to the time of the message flows in the proposed scheme. Fig. 4 was drawn according to the time of the message flows in Fig. 2. The total handover latency of [17] is

Table 1 System parameters and their descriptions. Notations

Descriptions

C mt C lm C mn C am C ad αt αm αl αa αd θ σ λi λo N mn Tf

The transmission cost between the MAG and MN The transmission cost between the LMA and MAG The transmission cost between the MAG and MAG The transmission cost between the AAA and MAG The transmission cost between the AAA and DNS The processing cost of location update at MN The processing cost of location update at MAG The processing cost of location update at LMA The processing cost of location update at AAA The processing cost of location update at DNS The active mode rate The ratio of the MN becoming active mode The rate of session arriving at the MN The rate of the MN sending packets Total number of mobile nodes in a subnet The average time each MN stay in each subnet

The total handover latency of the proposed scheme in this paper is T HL2 ¼ T link  switching þT RS þ T AAA  Auth þ T P  registration þ T RA ð2Þ The proposed scheme shortens the total handover latency by T DNS  Update compared with that by [17]. The latency time is as below, until a changed address can be checked by a CN which tries to communicate with an MN, after a DNS update. T DL1 ¼ T link  switching þ T RS þT AAA  Auth þ T P  registration þ T DNS  Update

ð3Þ

In this proposed scheme, the latency time is as below until a CN can check a changed IP address. T DL2 ¼ T link  switching þ T RS þT AAA  Auth

ð4Þ

In the proposed scheme, the DNS update time is reduced by T P  registration þ T DNS  Update compared with that by [17]. This is possible because the AAA replaces the role of the LMA, which updates the DNS. Therefore, at the same time, the LMA receives an AAA Query without the operations of the PBU and PBA, and it can update the DNS. 5.2. Location update cost In D-PMIPv6, assume each MN may move randomly between N subnets and there are k subnets within a PMIPv6 domain. So, MNs will move out to the other subnets with equal probability 1/(N-1). In our model, we call the action of each MN moving out of a subnet “a movement”. At movement I, MNs may reside in a subnet. At movement 2, MNs may move to N-1 other subnets. We assume each MAG is a subnet and each PMIPv6 domain is a paging area. The probability of moving out of a PMIPv6 domain at movement M is as below:
m  2 N k N k Pm ¼ ; 2rmr1 ð5Þ N1 N1 It can be shown that the expectation of M is

T HL1 ¼ T link  switching þT RS þ T AAA  Auth þ T P  registration þ T DNS  Update þ T RA

5

1

ð1Þ

EðMÞ ¼ ∑ mpm ¼ 1 þ m¼2

N1 N k

ð6Þ

5.2.1. The location update cost of the active MN When an MN in active mode moves within subnets in the same PMIPv6 domain, the MAG sends a PBU message to the LMA to update the MN's location. The location update cost includes the transmission cost between network entries and the processing cost at the MN, MAG, and LMA. It is shown as follows: C ul ¼ 2C mt þ4C lm þ2C am þ 2αt þ 6αm þ 2αl þ 2αa

ð7Þ

When an MN in active mode moves between PMIPv6 domains, the MN's location information of LMA and DNS should be updated. Thus, the location update cost is: C ud ¼ 2C mt þ 4C lm þ 2C am þ 2αt þ 6αm þ2αl þ αa þ αd

ð8Þ

In D-PMIPv6, the MN moves out of a PMIPv6 domain at movement M. That is, the mobile node moves within the same PMIPv6 domain for (E[M]-1) movements. Suppose

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

6

Fig. 3. Handover latency of PMIPv6 with the existing scheme.

the average time that each MN stays in each subnet before making a movement is T f . Thus, the average location update cost per unit time for active mobile nodes is as below C active ¼

ðE½M   1ÞC ul þC ud E½M T f

ð9Þ

If an MN does not send and receive data for a long time, the MN's state will change from the active mode into the idle mode. Then, the MAG sends idle mode notification messages to the other MAG in the same PMIPv6 domain. Therefore, the cost is shown as follows: Ca  i ¼

σ ðk 1ÞC mm Tf

ð10Þ

From Eq. (11), it is clear that:

σ¼

ðλi þ λo Þθ ð1  θÞ

ð12Þ

where λi is the rate of sessions arriving at the MN, λo is the rate of the MN sending packets, and θ is the active mode rate. 5.2.2. The location update cost of the idle MN In the previous section, we assumed that a PMIPv6 domain is a paging area. So, when an MN in idle mode moves between subnets in the same PMIPv6 domain, the MAG does not send a PBU message to the LMA, but the MN needs to send a router solicitation message. Thus, the cost is: C q ¼ C mt þ αm þ αt

ð13Þ

where a is the probability of the MN's state change from active to idle. Assume that the network is in a dynamic equilibrium state, so the number of MNs changing from active to idle is equal to the number of MNs changing to active from idle per unit time. Based on a Markov chain, the balance equation is as follows:

When an MN in idle mode moves between PMIPv6 domains, the MN's location information should be updated in the LMA and DNS. The location update cost is the same as the cost of an active MN. Thus, the average location update cost per unit time for idle mobile nodes is:

σ ð1  θÞN mm ¼ ðλi þ λo ÞθN mm

C idle ¼

ð11Þ

ðE½M  1ÞC q þ C ud E½MT f

ð14Þ

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

7

Fig. 4. Handover latency of proposed scheme with the proposed scheme.

If an idle MN sends or receives data packets, the MN's state turns into the active mode, and then the serving MAG sends a PBU message to the LMA. The cost is Ci  a ¼

ðλi þ λo ÞC ul Tf

ð15Þ

The ratio of the active MN is θ. Thus, in D-PMIPv6, the location update cost for all the MNs is C d  pmip ¼ θðC active þ C a  i Þ þ ð1  θÞðC idle þC i  a Þ

ð16Þ

5.3. Paging cost 5.3.1. The cost of paging active MN In D-PMIPv6, a CN that wants to communicate with an MN requests the address of the MN from the DNS. The CN with the MN's address sends data packets to the MN via the LMA. The LMA knows that the MN is in active mode and then checks the binding cache entry to find the MAG to which the MN attaches. The complexity of the lookup is proportional to the total number of MNs in the database. Assume that there are N mn MNs in a MAG subnet and a

PMIPv6 domain consists of k MAGs. Therefore, the lookup cost is proportional to kN mn , and the weighting factor is α. Because the MN is in active mode, the paging cost is equal to the lookup cost. It is shown as follows:   P active ¼ θ kλi akNmn ð17Þ where λi is the average packet arrival rate for each MN, and θ is the ratio of the active MN.

5.3.2. The cost of paging idle MN When a CN wants to communicate with an MN and the LMA knows that the MN is in idle mode, the lookup cost in the LMA is also proportional to the total number of MNs. Then, the LMA sends paging request messages to all the MAGs controlled by the LMA in the same PMIPv6 domain. This kind of signalling cost is proportional to the number of MAGs, the average transmission cost between the MAG and the LMA, and the average packet arrival rate for each MN. On receiving the paging request message, the serving MAG performs binding cache entry lookup, and the cost is proportional to the number of MNs in a MAG subnet. Thus,

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]] 30

60

25

50

20

Total Signalling Cost

Total Signalling Cost

8

Tf=5 Tf=15 Tf=50

15 10 5

40 30 20 10

0 0

5

10

15

20

25

30

The PMIPv6 Domain Size, k

0 0

5

10

15

20

25

30

The PMIPv6 Domain Size, k

Fig. 5. Total signalling cost vs. PMIPv6 domain size (k), average residence time.

Fig. 6. Total signalling cost vs. PMIPv6 domain size (k), session arrival rate.

the cost of paging an idle MN is:   P idle ¼ ð1  θÞ kλi ðakN mn Þ þ λi ðaN mn Þ þ kλi kC lm

reduces the influence of a high mobility rate on the location update cost. So, we can conclude that a larger PMIPv6 domain under high mobility rate accompanies the optimal operation. As shown in Fig. 5, the total signalling cost increases as the average residence time decreases. This is because when T f is small, the mobility rate is high, which leads to frequent location updates. Due to the fixed session arrival rate, the paging cost is constant. Thus, the total signalling cost including the location update cost and the paging cost increases as T f decreases. Next, we study how the total signalling cost varies with the size of the PMIPv6 domain under different session arriving rates. We set T f to 15 and θ to 0.1. Fig. 6 indicates that when λi is large, the total cost in the small PMIPv6 domain is less than the one in the large PMIPv6 domain. This is because a high session arriving rate leads to excessive paging request messages. Therefore, with the PMIPv6 domain size, increasing the paging cost dominates, and the total cost rises rapidly. Nevertheless, when λi is small, as the PMIPv6 domain size increases, the location update cost decreases faster than the paging cost increases, which leads to a rise in the total cost. Note that the total signalling cost increases as the average packet arrival rate increases. The reason is that when λi is large, paging request messages are often sent, which brings about high paging signalling cost. In summary, as the PMIPv6 domain size increases, the location cost decreases, but the paging cost increases. Under high mobility rate and low session arriving rate, the optimal performance can be achieved with a larger PMIPv6 domain. In addition, regardless of the PMIPv6 domain size, the total cost decreases as the average residence time increases, or the average packet arrival rate decreases.

ð18Þ

5.4. The total cost By summing up the location update cost (16) and the paging cost (17) and (18), we obtain the total signalling cost as follows: C total ¼ C d  pmip þ P active þ P idle

ð19Þ

6. Numerical result Based on the above analysis, in this section, we give the analytical performance results of the total signalling cost. We set the system parameter values as follows: pffiffiffi N ¼ 30; C mt ¼ 6; C lm ¼ C mm ¼ C am ¼ k; C ad ¼ 15; αt ¼ 5; αm ¼ 10; αl ¼ 20; αa ¼ 20; αd ¼ 50; N mn ¼ 10; α ¼ 0:1

6.1. The impact of the PMIPv6 domain size on the total signalling cost First, we confirm how the size of the PMIPv6 domain affects the total signalling cost changes under different average residence times. In this experiment, we set λi and λo to 0.002 and θto 0.1. Other parameters are set to the default values shown in Table 1. Fig. 5 illustrates that if T f is large, the total signalling cost increases with the PMIPv6 domain size. The increasing PMIPv6 domain size accompanies a low mobility rate, which reduces the update cost and increases paging cost. There is a reason why the paging cost dominates. Whenever paging occurs for an idle MN, paging request messages are sent to all the MAGs in the same PMIPv6 domain. The larger the PMIPv6 domain size is, the more the cost of paging request messages is. Thus, the total cost increases as the PMIPv6 domain size increases under a low mobility rate. In contrast, when T f is small, a large PMIPv6 domain size

6.2. The impact of the active mode rate on the total signalling cost In this section, we investigate how the total signalling cost varies with the active mode rate. Let λi ¼ λo ¼0.01, T f ¼ 10, and θ varies from 0.1 to 1. Fig. 7 illustrates the total rate under different PMIPv6 domain sizes. The larger the active mode rate is, the more the location update cost is, but the paging cost is smaller.

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

34 32

Total Signalling Cost

30 28 26 24 22 20

k=10 k=15 k=20

18 16 14 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

The Active Mode Rate

Fig. 7. The total signalling cost vs. the active mode rate. 50

Total Signalling Cost

40

30

20

10

0 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

The Active Mode Rate

Fig. 8. The total signalling cost vs. the active mode rate. 20

Total Signalling Cost

18

λi=0.0005 λi=0.005 λi=0.02

16 14 12 10 8 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

The Active Mode Rate

Fig. 9. The total signalling cost vs. the active mode rate.

When the PMIPv6 domain size is small, with the active mode rate increasing, the rising speed of the location update cost is faster than the decreasing speed of the paging cost. Therefore, the total signalling cost rises as the active mode rate increases. However, if the PMIPv6 domain size is large, the total signalling cost decreases as the active mode rate increases. Thus, we could design the size of the PMIPv6 domain depending on different active rates. Fig. 8 shows how the total signalling cost varies with the active mode rate under different residence times. Assume k¼10, λi ¼ λo ¼0.01, and T f ¼5, 15, 50. When T f is small, the MN stays in the subnet for a short time. As a result, there is an evident increase of the location update cost. Thus, it results in a total cost increase as the active

9

mode rate rises. However, if T f is large, the paging cost will reduce more rapidly than the location update cost does. Consequently, the total signalling cost decreases as the active mode rate increases. Fig. 9 describes the total signalling cost changes with the active mode rate under different session arrival rates. Assume k ¼10 and T f ¼15. When λi is large, the total cost decreases as the active mode rate rises. This operation happens because a high session arriving rate leads to excessive paging request messages, causing dominant paging cost. As the active mode rate increases, the number of the active MN does as well, but the amount of the idle MNs decreases. Consequently, the message number of paging for an idle MN reduces faster than that of the location update, which results in the reduction of the total cost. Nevertheless, when the session arriving rate is small, there are fewer paging messages compared to location update messages. With the active mode rate increasing, the location update cost increases faster than the paging cost decreases, so the total cost increases. In summary, as the active mode rate increases, the location cost increases, but the paging cost decreases. In the case of small PMIPv6 domain size, high mobility rate, and low session arriving rate, the total cost decreases as the active mode rate decreases. 7. Conclusion We have proposed a location management scheme for PMIPv6 based on DNS. In the proposed scheme, we consider the DNS as a location manager to support global mobility in PMIPv6. In addition, a paging extension scheme was introduced to PMIPv6. To evaluate the location management scheme, we established a theoretical model and formulated the location update and the paging cost to analyse the performance. The performance results show that there is a trade-off between the location cost and the paging cost. As the PMIPv6 domain size decreases, or the active mode rate increases, the location cost increases but the paging cost decreases. In the case of high mobility rate and low session arriving rate, the optimal PMIPv6 domain size is large. Under small PMIPv6 domain size, high mobility rate, and low session arriving rate, the optimal performance can be achieved with a low active mode rate.

Acknowledgements This research was supported by the Ministry of Trade, Industry and Energy (MOTIE), through the Special Education program for Industrial Convergence. Also, this research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF2010–0024695). This article is a revised and expanded version of a paper entitled ‘Performance Analysis of DNS-assisted Global Mobility Management Scheme in Cost-Optimized Proxy Mobile IPv6 Networks' presented at the International

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i

10

C. Cho et al. / Information Systems ] (]]]]) ]]]–]]]

Symposium on Advanced and Applied Convergence held from November 14 to 16, 2013,in Seoul, Korea. Appendix A. Supplementary information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.is. 2014.08.001.

References [1] S. Gundavelli, K. Leung, V. Devarapalli, K. Chowdhury and B. Patil, Proxy Mobile IPv6 IETF RFC 5213, 2008. [2] J. Kempf, Goals for Network-Based Localized Mobility Management (NETLMM) IETF RFC 4831, 2007. [3] Ki-Sik Kong, Wonjun Lee, Youn-Hee Han, Myung-Ki Shin and HeungRyeol You, mobility management for all-IP mobile networks: mobile IPv6 vs. Proxy Mobile IPv6 IEEE Wireless Commun., 2008. [4] G. Giaretta, Interactions between PMIPv6 and MIPv6: scenarios and related issues IETF draft, draft-ietf-netlmm-mip-interactions, 07, 2010. [5] A.A.S. Reaz, M. Atiquzzaman, Fu Shaojian, Performance of DNS as location manager for wireless systems in IP networks, in: Proceeding of the IEEE Global Telecommunications Conference (GLOBECOM), November 2005. [6] Jong-Hyouk Lee, Tai-Myoung Chung, Sangheon Pack, S.r.i. Gundavelli, Shall we apply paging technologies to proxy mobile IPv6, MobiArch'08, 2008. [7] Ian F. Akyildiz, Joseph S.M. Ho, Yi-Bing Lin, Movement-based location update and selective paging for PCS networks, IEEE/ACM Trans. Netw. 4 (4) (1996). [8] Xiaowei Zhang, Javier Gomez Castellanos, Andrew T. Campbell, PMIP: paging extensions for mobile IP, Mob. Netw. Appl. 7 (2002). [9] S. Bellovin, Using the DNS for system break-ins, in: Proceedings of the Usenix Security Symposium, 1995. [10] Jong-HyoukLee, Joong-Hee Lee and Tai-Myoung Chung, An adaptive inter-mobility support scheme for NetLMM, in: Proceedings of the 2nd International Conference on Systems and Networks Communications (ICSNC), August 2007. [11] H. Soliman, C. Castelluccia, K. El Malki and L. Bellier, Hierarchical mobile IPv6 (HMIPv6) mobility management, IETF RFC 5380, 2008. [12] Domain Name System Security Extensions, 〈http://www.ietf.org/rfc/ rfc2535.txt〉, March 1999. [13] DNS Security Introduction and Requirements, 〈http://www.ietf.org/ rfc/rfc4033.txt〉, March 2005. [14] Resource Records for the DNS Security Extensions. http://www.ietf. org/rfc/rfc4034.txt, March 2005. [15] Protocol Modifications for the DNS Security Extensions, 〈http:// www.ietf.org/rfc/rfc4035.txt〉, March 2005.

[16] Jiang Xie, Akyildiz I.F. A distributed dynamic regional location management scheme for mobile IP, in: Proceedings of the IEEE INFOCOM, June 2002, pp. 1069–1078. [17] Feng Qiu, Xiaoqian Li, Hongke Zhang, A location management scheme based on DNS in proxy mobile IPv6,in: Proceedings of the Future Information Networks (ICFIN), October 2009, pp. 38–44.

Chulhee Cho received his B.S. degree in Department of Physical Education from the HanYang University, Seoul, Republic of Korea, in 1992 and his M.S. degrees in Information Security from the Sungkyunkwan University, Seoul, Korea, in 2013. He has worked in the field of IT Strategy & IT Development from 1997 to 2008. He has been the general manager in charge of IT Security of a public corporation from 2009 to 2013, respectively. His research interests include mobile computing, mobility management for vehicular networks, cryptography, network security, protocol operation based performance analysis and Internet security. He is a member of the IEEE, KIISC, KSII, and KIPS.

Jeong-Jin Kang is currently a member of the faculty of the Department of Information and Communication at the Dong Seoul University in SeongNam, Republic of Korea since 1991, and currently the President of the Institute of Internet, Broadcasting, and Communication(IIBC). During 3 years from February 2007 to February 2010, he worked as a visiting professor at the Department of Electrical and Computer Engineering, the Michigan State University. He was a lecturer of the Department of Electronic Engineering at the (Under) Graduate School (1991–2005), the Konkuk University. Dr. Kang is a member of the IEEE Antennas and Propagation Society (IEEE AP-S), the IEEE Microwave Theory and Techniques Society (IEEE MTT-S), and a life member of the Institute of Internet, Broadcasting, and Communication (IIBC), Republic of Korea. His research interests involve smart mobile electronics, RF mobile communication, smart convergence of science and technology, RFID/ USN, u-Healthcare, and new media service.

Jongpil Jeong received his B.S. degree in engineering from the Sungkyunkwan University and his M.S. and Ph.D. degrees in computer engineering from the Sungkyunkwan University, Suwon, Korea, in 2003 and 2008, respectively. He was a Research Professor with Sungkyunkwan University in 2008–2009 and 2011, and a visiting professor with the Department of Interaction Science (WCU Programme) in the Sungkyunkwan University from 2009 to 2010. He started his academic profession at the Research & Business Foundation of Sungkyunkwan Univeristy, Korea in 2012 as an assistant professor. He received Excellent Research Awards, twice, from the Department of Electrical and Computer Engineering, Sungkyunkwan University, Republic of Korea (2007) and from the KSII (Korea Society for Internet Information), Korea (2011). His research interests include mobile computing, mobility management for vehicular networks, interaction science, sensor networking, protocol operation based performance analysis, Internet security, MIPv6 and ubiquitous computing. He is a member of the IEEE, KIISC, KSII, KIPS, and IEEK.

Please cite this article as: C. Cho, et al., Performance analysis of DNS-assisted global mobility management scheme in cost-optimized proxy mobile IPv6 Networks, Information Systems (2014), http://dx.doi.org/10.1016/j.is.2014.08.001i