Efficient Handoff using Mobile IP and Simplified

0 downloads 0 Views 582KB Size Report
Dec 8, 2004 - In those two solutions, mobile IP is used as a handoff mechanism ..... Forouzan B A (2003), TCP/IP Protocol Suite, McGraw-Hill, 726-727.
Presented at GNSS 2004 The 2004 International Symposium on GNSS/GPS Sydney, Australia 6–8 December 2004

Efficient Handoff using Mobile IP and Simplified Cellular IP S. Omar

School of Surveying & Spatial Information Systems The University of New South Wales, UNSW, Sydney, NSW 2052, Australia Email: [email protected]

J. Lee

School of Computer Science & Engineering The University of New South Wales, UNSW, Sydney, N.S.W 2052, Australia Email: [email protected]

Joon ho Lee ABSTRACT Integration of wireless LAN and cellular network will possibly bring ubiquitous internet access. This study states several problems of currently suggested integration techniques and provides proposals to overcome these shortcomings. A current cellular IP will be simplified to improve the handoff performance which will be integrated with mobile IP. The simulator has been developed instead of a test-bed and the test result will be compared with the previous studies. Also, well-known simulator such as NS-2 will be used for comparison. KEYWORDS: Wireless Integration, Roaming, Simplified Cellular IP, Mobile IP Simulation

1. INTRODUCTION Wireless networks have been growing dramatically in the last decade. Mobile devices such as PDAs, laptops and mobile phones are widely used all around the world. These devices allow users to connect to the internet more conveniently. Wireless local area networks are also growing. Several places have wireless LAN infrastructure and these places are called Hotspots (Balachandran and VoelKer, 2003). These places include airports, libraries, university campuses, cafés, offices, hotels, shopping malls and so on. It is stated that by the year 2007, these hotspots areas will increase to 20,000 places. In the near future, this will make it possible to access the internet anywhere, anytime. However, each wireless network has several shortcomings. For example, wireless local area network such as WiFi hotspots provides fast data transfer speed but its coverage area is limited to several meters (Henry, 2002). According to Henry, it states 4 shortcomings to overcome for WiFi (2002). WiFi should be improved for ease of use, security, mobility and

network management. Cellular networks have been widely used for mobile phones. The obvious advantage of cellular networks is that it covers a wide range area. For cellular networks such as cdma2000, it has wide area coverage but does not provide fast data transfer. Thus, by integrating these wireless techniques it will allow users a fast and wide coverage access. There have been several studies for integrating wireless local area and cellular networks (Buddhikot et al, 2003). According to Buddhikot, two solutions have been proposed for the integration of cdma2000 and wireless LAN. These are tightly coupled and loosely coupled architecture. A tightly coupled architecture makes a link between wireless LAN and core 3G network. A gateway router of wireless LAN connects to PDSN. The gateway hides WLAN information and all the 3G protocols needs to be implemented in the gateway. Furthermore, a mobile node also needs to implement all the 3G protocols on top of its standard wireless LAN protocol. However, this architecture has several disadvantages. Since wireless LAN is directly connected to the core 3G network, the same operator needs to own both the cellular network and the wireless LAN. Secondly, an additional implementation on mobile nodes is required. As it was mentioned before, all the 3G protocols have to be implemented on top of a standard wireless protocol with mobile nodes. Loosely coupled architecture connects the gateway router to the global internet. Thus, there is no direct link between the cellular network and wireless LAN network. This makes it possible to separate the data paths in the WLAN and 3G. Several different protocols could be used but to successfully interoperate, it requires an authentication protocol between the two networks to be the same. The advantage of a loosely coupled approach is that it allows independent deployment and traffic engineering of WLAN and 3G networks. Furthermore, if there is a roaming agreement with several partners, wider areas could be covered by this integration. Due to the greater advantage of loosely coupled over tightly coupled approach, this research paper will be based on a loosely coupled integration as shown in Figure 1.

Figure 1. Overall Architecture

In those two solutions, mobile IP is used as a handoff mechanism between cdma2000 and WLAN. Mobile IP is the one of the most popular internet protocols for mobile technology. It consists of three functionalities. These are agent discovery, registration and tunnelling (Perkins, 1997). The agent discovery involves mobility agents advertising their availability on each link for which they provide service. When a mobile node is away from its home, it should register its care-of-address to its home. This is called registration. A tunnel is created between a home agent and care-of-address when a mobile node is in a foreign network. In loosely coupled architecture, when a mobile node moves into a wireless LAN cell, a tunnel between the home agent and the gateway router in WLAN will be created. When the mobile node moves out of the WLAN, a tunnel between the PDSN will be created if the mobile node is in a different PDSN domain (Buddhikot et al, 2003). This approach is a good handoff technique between cdma2000 and WLAN. However, the disadvantages of mobile IP have been reported (Perkins, 1997). Those are traffic overload and packet lost during handoff processes. Although several solutions that overcome limitations of mobile IP have been examined, it is still not enough to perform seamless handoff (Aust et al, 2002). The implementation of this study is based on the previous studies and aimed to improve on them. The structure of this article is as followed; the problems of current mobile IP, technical details to overcome those problems, evaluation result of the simulator followed by conclusion remarks 2. RELATED WORK In previous work, an alternative architecture has been proposed to improve handoff performance (Omar and Lee, 2004). This is called hierarchical arrangement of foreign agents. According to Perkins, this proposal has been reported. A foreign agent in WLAN could be established hierarchically and multiple foreign agents will have an advertisement procedure at the same time. It will make registration procedure localized to WLAN. Figure 2 shows this architecture. For example, while a mobile node is in FA7 it receives an agent advertisement for FA4, FA2 and FA1. Its home agent believes that the mobile node is located in care of address FA1. FA1 believes that the mobile node is located in care of address FA2 and so on. When the mobile node moves to FA8, a hierarchical registration has to occur up to FA4 and the rest of the tunnels remain the same. When the mobile node moves to FA9, the mobile node received an agent advertisement of FA9, FA6, FA3 and FA1. It will cause the hierarchical registration procedure up to FA1 by comparing the previous and current lineage. A tunnel from PDSN to gateway router will remain the same and only tunnels between lower level foreign agents have to be re-established. This architecture will allow users for fast handoff because registration messages need to travel only within WLAN. Furthermore, a home and foreign agents in WLAN are relatively closer to each other, it will reduce packet lost and traffic overload during registration procedure.

Figure 2. Hierarchical Arrangement of Foreign Agents

3. PROBLEM STATEMENT Several integration architectures have been proposed (Buddhikot et al, 2003). Some of these studies integrate GPRS with a wireless LAN (Calvagna et al, 2003). The others integrate cdma2000 with a WLAN (Buddhikot et al, 2003). All of these suggestions are based on Mobile IP. Mobile IP is a reasonable protocol for wide area coverage with less frequent handoff users. However, there are several issues about mobile IP. For example, if a mobile node wants to send packets to another node in the same network, packets have to cross to home agents and back to foreign agents. This creates erroneous network communication as a direct local link could have been used which would be much more efficient. Furthermore, when the remote host communicates with a mobile host that is not attached to the same network as the mobile host, triangular routing possibly occurs. In triangle routing, a mobile host can send packets directly to a remote host but a remote host has to send its packets to the home address first before it is rerouted back to the mobile node. Thus, packets travel two side of triangle instead of one (Forouzan, 2003). Furthermore, mobile IP requires registration information after each handoff for each mobile node. A home agent possibly exists relatively far from a foreign agent and this distance could increase handoff latency and load on the traffic (Valko, 1999). Frequent migration also generates too much traffic between the visited and home network during the registration process. This traffic can hamper the speed and efficiency of the network. In addition, if the mobile user is idle, the update message is sent to the home agent. This load is proportional to the number of mobile hosts and not to generated traffic (Valko, 1999). Thus, mobile IP does not perform an efficient handoff in small areas. Furthermore, according to the work of Buddhikot, because mobile IP could make a tunnel directly to a mobile node when a user is within WLAN, it requires mobile nodes to support mobile IP functionalities. This makes extra implementation

on mobile nodes. The cellular IP has been introduced to overcome the limitations of mobile IP. Although current cellular IP specification improves the handoff performance, problems still exist. Since the cellular IP routers need two caches to store the status of mobile users, it will require more memory usage. Also, to keep paging and routing cache up-to-date, a mobile node is responsible for sending update packets to its routers. Thus, the mobile node should update its status in the cellular IP router as well as the mobile IP router. This increases traffic overload. 4. PERFORMANCE IMPROVEMENT OF MOBILE IP The inefficiency of mobile IP in a micro area could be improved by localizing handoff procedures within a WLAN. Two different IPs for different coverage will be used. Cellular IP will handle micro areas and mobile IP will handle macro areas. Also, cellular IP will be improved by simplifying caches and update packets. The technical details of these methods will be provided in next sub-sections. 4.1 Integration of Mobile IP and Cellular IP

An integration of mobile IP and cellular IP will be based on loosely coupled approach as it mentioned before. Figure 3 shows abstraction of integration. If a user moves into a wireless local network area, mobile IP will perform tunnelling between the PDSN and the gateway router. Likewise, when a user moves out of the WLAN area, mobile IP will tunnel to PDSN if that network is foreign. As long as the user is within the WLAN network, the mobile IP tunnel remains the same and all the packets are forwarded to the gateway router and within WLAN, cellular IP will be used to perform handoff. Therefore, mobile IP will handle global area handoffs and cellular IP will handle local area handoffs.

Figure 3. Abstraction of integrating Mobile IP and Cellular IP

Cellular IP has been proposed to overcome the inefficiency of mobile IP in micro areas (Valko, 1999). The design goals of the cellular IP are simplicity, scalability and performance scalability. Cellular IP can be implemented on top of regular IP routers (simplicity). Its distributed location management makes it possible to use the same protocol and the same topology-unaware nodes (scalability). Lastly, it allows obtaining locally available level of

service (performance scalability). Cellular IP is fully compatible with IP and keeps two caches. These are paging and routing caches. Not all the routers have those caches. Thus, if a router does not store any caches, it will forward the received packets to all the connected routers and nodes. Paging cache is used for idle users. When a user is in an idle state, the mobile node keeps generating paging-update packets. This packet will be sent to the nearest base station. The paging-update packet will travel to the gateway router on a hop-by-hop basis. If any router with paging cache receives this packet, it will update its paging cache. When the idle host moves, it keeps sending paging-update packets to update their location. Mappings in paging cache will be timed out in several minutes depending on its implementation. For active users, it creates a route-update packet. This packet will be sent to its base station and it will travel to the gateway router similar to a paging-update packet. Routing cache in the routers will be updated. Mapping in routing cache will be timed out shorter than paging cache. By using these two caches, handoff could be performed efficiently. Handoff is always initiated by the mobile node. When users move into a new access point, packets will be redirected from an old station to the new station. The first packet will automatically configure a new path for the routing cache. Before the timeout occurs in the routing cache, packets will be delivered to both old and new stations to prevent packet loss. Once a mapping in the routing table gets timed out, the packets will be forwarded to the new station only. Figure 4 illustrates this scenario. When a user moves from station F to station D, while data is transmitting, it will initialize handoff to station C. Assuming that only station C and E contains routing tables, a path in station C will be updated. Before mapping in C timed out, station C will send packets to both station E and D. Once routing cache gets timed out, it will send packets to station D only. Since cellular IP is used within WLAN, any migration within the WLAN is transparent to mobile IP and the tunnel is shared by all users (Perkins, 1998). It reduces redundant registration within WLAN. This separation of global and local area coverage will improve handoff performance. Since coverage of wireless LAN access points is far smaller than coverage of base stations, users are likely to move between cells more frequently. Thus, using mobile IP within WLAN would cause a dramatic traffic overload for registration.

Figure 4. Cellular IP Handoff Scenario

Using two different IPs for different area ranges will optimize handoff performance. Because a tunnel is kept remaining, the registration process is not required for handoff within

WLAN. Cellular IP will handle more efficient handoff within WLAN. This also makes packet lost and traffic overload to be minimized. More users will be allowed access in the WLAN as well. 4.2 Simplified Cellular IP

In normal cellular IP specification, a mobile node sends its routing update packet or paging update packet to update its routing cache. Likewise, the mobile node is also responsible for updating its mobile IP registration information in foreign and home agents when it is outside of home network. As mentioned in the previous sub-section, mobile IP tunnel is shared by all users. Although the tunnel is shared, it is required to transmit both an update packet for cellular IP and a registration packet for mobile IP. Therefore, it will reduce traffic overhead if only one packet could update both routing cache and registration cache. In this simplified cellular IP, the registration message of mobile IP will be used to update both cellular routers and mobile routers. When this message gets sent from the mobile node, it will pass through the cellular IP router in a hop-by-hop base and update its cache. When this message reaches the gateway router which operates mobile IP functionality, it will be used as the registration message for mobile IP. In addition, instead of having two caches, only one cache will be used to store mobile user's status. Thus, whether the mobile users are active or not, their information will be stored in the same cache. 5. EVALUATION The simulator is used for measuring handoff performance. This simulator is implemented in Java and operates on Intel Pentium 1500Mhz processor. The evaluation result will be divided into three parts. In first subsection, it provides average memory usages of cellular IP router cache. In second subsection, it introduces average update packets sent by user. Then, the analysis of packet loss during handoff procedure is described in the third part. 5.1 Average Memory Usage of Cellular Routers

The simplified cellular IP has been simulated in comparison of normal cellular IP router. The evaluation result is shown in Figure 5.

Figure 5. Average Memory Usage of Cellular Routers

It shows that if the registration packets get sent more frequently than the routing update packet, the routers require less memory space than normal cellular IP routers. In addition, if the packets get sent less frequently than the paging update packet, it requires more memory space than normal cellular IP routers. For example, it shows that in case of normal cellular IP routers, average of 18.13 bytes is used. However, in simplified cellular IP, it requires the average of 19.36 bytes for time interval of 20 seconds and 10.83 bytes for 5 seconds of the registration messages. For time interval of 20 seconds, the simplified cellular IP requires more memory. However, if time interval gets shorter, it requires less memory spaces than normal cellular IP routers. It shows that it reduces 40%, 23% and 9% of memory spaces with time interval of 5, 10 and 15 seconds respectively. This result is expected because if the registration messages get sent less frequently, each entry in the cache should live for a long time since it is assumed that time interval of the registration packet and TTL time in the cache are the same. Thus, if the mobile node has moved to another network, the entries will still exist and get removed later. Likewise, if the packets get sent more frequently, the cache will be updated as soon as the mobile node moves away. However, too much packet will cause heavy traffic overhead so that this result should be considered together with the evaluation of traffic overhead 5.2 Traffic Overhead

The evaluation result of traffic overhead is shown in Figure 6. The simulator shows that if packets get sent every 5 seconds, it will cause traffic overhead up to 8 bytes per second whereas normal cellular IP will cause 2.93 bytes per second for idle user and 7.6 bytes per second for active user.

Figure 6. Traffic Overhead

This also gives obvious result that the more packets are sent, the more traffic overheads are created. However, by comparing the result of previous subsection, it is possible to achieve the optimized time interval. In both simulations in section 5.1 and 5.2, it is assumed that the paging update packets get sent at the interval of every 15 seconds and the routing update packets get sent at every 5 seconds. It is proved that if the registration packets get sent in between 5 and 15 seconds, the memory will be saved and the traffic overhead will be reduced. For example, it requires 2.00

bytes and 3.60 bytes for time interval of 5 and 15 seconds respectively. This is 31% reduction compare to idle host and 73% reduction compare to active host for 5 seconds time interval. For 15 seconds time interval, it increases 18% of traffic overhead compare to idle user and reduces 52% compare to active user. Although there is an increase of traffic overhead when it compares to idle host, it will reduce the traffic overhead dramatically when the mobile node is active. Therefore, it could be concluded that the time interval of 5 and 15 seconds (between the interval of paging update and routing update packets) will reduce both traffic overhead and memory usage of routers. 5.3 Packet Loss

The packet loss is caused by several factors such as environmental clutter, mobile speed and cell coverage overlap (Campbell et al,2000). However, since this simulator is built in one single computer, it is unrealistic to measure packet loss. Therefore, previous studies are used to approximate its packet loss during handoff procedure. According to Edwards and Suryakumar, it shows that the faster the mobile node moves, the more packets will be dropped. It shows that the mobile speed of 10m/s could loss 0.05% of packets and the mobile speed of 40m/s could loss 0.06% of packets. For the cell coverage overlap case, it is shown that the minimum overlap coverage is required for seamless handoff (Edwards and Suryakumar, 2003). It shows that the overlap area of 0 could drop 0.06% of packets and the overlap coverage region of 30m could drop 0.05% of packets. By comparing those results with the simplified cellular IP, it is approximated that the result will be the same. The simplified cellular IP simplifies the cache update procedure and uses only one cache. This is not a factor to cause packet loss. Therefore, the packet loss of normal cellular IP and the simplified cellular IP will be the same. 5.4 NS-2 Simulation

NS-2 simulation is used to compare this work with previous works. Columbia university team made a NS-2 extension for mobile functionalities. Thus, this NS-simulation is based on ns-2 version 2.1b6 with CIMS (Columbia IP Mobility Software) extension pack and this extension packet uses normal cellular IP specification. The previously released article is also used for measuring cellular IP performance (Campbell et al, 2002). The simulation has been performed in network topology shown in Figure 7. According to Campbell, in this topology, each router is acting as a gateway to the internet. All the wired connections is modelled as a 10Mb/s duplex link with 2ms delay. Mobile hosts connect to access points using the NS-2 carrier sense multiple access with collision avoidance wireless link model where each access point operates on a different frequency band. In this simulation, single mobile host is used with UDP packet. This host will move between access points in different speed and it always go through maximum overlapping area.

Figure 7. NS-2 Simulation Network Topology (Campbell et al, 2002)

A constant speed rate of 20m/s and an overlapping area of 30m have been used for simulation (Campbell et al, 2002). This simulation has been performed in three different scenarios with various numbers of hops between the crossover node and access point (crossover distances). For example, according to Figure 7, crossover distance of AP1 and AP2 is 1 because there is only one hop (R3). Thus, the crossover distance will be 1, 2 and 3 for AP1-AP2, AP2-AP3 and AP3-AP4 respectively. This simulation result is shown in Figure 8. It shows that if the crossover distance is larger, the packets will be delayed longer. According to Campbell, it will cause an extra packet delay of 2ms for each additional hop.

Figure 8. Cellular IP UDP packet performance result (Campbell et al, 2002)

The handoff performance was measured using TCP packets as well (Campbell et al, 2002).

The simulation result is shown in Figure 9. This dotted line shows sequence number of data packet received by the mobile host. In this simulation, at 14.75s the cellular IP handoff procedure occurs. The TCP timeout causes data packet loss so no data packet is transmitted during the timeout period. This results in a serious degrading of TCP performance.

Figure 9. TCP Sequence Number at the time of a Cellular IP (Campbell et al, 2002)

As mentioned before in the previous subsection, the simplification of cellular IP specification will not affect the packet loss. Thus, the packet loss of normal cellular IP and simplified cellular IP will be the same. However, it shows that crossover distance is also another factor to cause packet loss. It shows that minimizing the hop count will reduce the delay. Therefore, careful arrangement will minimize the packet delay and it will make minimum packet loss. 6. CONCLUSION The integration of the cellular network and wireless LAN will give great advantages. It will provide wide area coverage with fast data transmission rate by using cellular and wireless local area network. The provided technique (simplified cellular IP) has been used for integration of cellular network and wireless LAN. The suggested simplification of cellular IP has been improved current cellular IP specification. It reduces the memory usage in routers and the traffic overhead by carefully adjusting the time interval of the registration message. It reduced 40%, 23% and 9% of memory spaces with time interval of 5, 10 and 15 seconds respectively. For traffic overhead, it requires 2.00 bytes and 3.60 bytes for time interval of 5 and 15 seconds respectively. This is 30% reduction compare to idle host and 73% reduction compare to active host for 5 seconds time interval. This shows that the registration message should be sent in between 5 and 15 seconds. This is during paging update and routing update packet time interval. However, the packet loss will be approximately the same as normal cellular IP since the simplification does not affect the factors of packet loss. Further research needs to be performed and also actual implementation should be considered to accurately measure the performance of the simplified cellular IP.

REFERENCES Balachandran A, VoelKer G M (2003) Wireless hotspots: Current challenges and future directions, Journal of ACM Press, 1-9 Henry P S (2002) WiFi: What’s next?, IEEE Communications Magazine, 40(12): 66-72 Buddhikot M M, Chandranmenon G, Han S, Lee Y, Miller S, Salgarelli L (2003), Design and Implementation of a WLAN/cdma2000 Interworking Architecture, IEEE Communications Magazine, 41(11): 90 - 100 Calvagna A, Morabito G, La Corte A (2003), WiFi bridge: wireless mobility framework supporting session continuity, Proceedings of the First IEEE International Conference on, 23-26 March 2003, 79 - 86 Aust S, Protel D, Konsgen A, Pampu C, Gorg C (2002), Design issues of Mobile IP handoffs between General Packet Radio Service (GPRS) networks and wireless LAN (WLAN) systems, Wireless Personal Multimedia Communications 2002. The 5th International Symposium on, 27-30 Oct. 2002, 868 – 872 Perkins C E (1997), Mobile IP, IEEE Communications Magazine, 35(5): 84 - 99 Forouzan B A (2003), TCP/IP Protocol Suite, McGraw-Hill, 726-727 Valko A G (1999), Cellular IP: A New Approach to Internet Host Mobility, ACM SIGCOMM Computer Communication Review, 29(1): 50 - 65 Perkins C E (1998), Mobile Networking Through Mobile IP, IEEE Internet Computing, 2(1): – 69

58

Edwards G, Suryakumar N (2003), Cellular IP Performance, IEEE Wireless Communication and Networking, 3:2081-2085 Campbell A T, Gomez J, Kim S, Valko A G, Wan C Y, Turanyi Z R (2000), Design, Implementation and Evaluation of Cellular IP, IEEE Personal Communications, 7: 42-49 Campbell A T, Gomez J, Kim S, Wan C Y, Turanyi Z R, Valko G (2002), Comparison of IP Micromobility Protocols, IEEE Wireless Communications, 2 – 12 Omar S, Lee J (2004), Handoff Improvement between WiFi and CDMA2000, In proceedings of the International Conference on Communications in Computing, Las Vegas, Nevada, U.S.A, 21 – 24 June, 2004