Sie sind auf Seite 1von 9

A comparison of mobile IP handoff mechanisms

Luke Niesink l.d.j.niesink@student.utwente.nl ABSTRACT


After briefly explaining mobile IP and handoffs this paper introduces the problems that the original mobile IP specification has. The main problem considered is improving the smoothness of the handoffs. A lot of the solutions that have been suggested to solve these problems are briefly explained and finally compared. In conclusion it is established that a combination of the mechanisms discussed is the best approach in most cases. the mechanisms that will be discussed in this paper. In order to make a proper comparison we need to establish criteria, these will be explained in section 5. The comparison of the mechanisms on the basis of those criteria will be made in the 6th section, and finally the conclusions that can be drawn from all this will be elaborated on in section 7.

2. MOBILE IP
In the first part (2.1) of this section the problems concerning mobility in conventional routing mechanisms used by the internet, will be explained. These problems are the reason for creating mobile IP. What mobile IP is will be explained in section 2.2.

Keywords
Mobile IP, handoff latency, seamless handoff, hierarchical, fast handoff, route optimization, HMIPv6

1. INTRODUCTION
In IP based networks whenever you switch to a new router, for instance in a WLAN environment, you are assigned a new IPaddress. Because of the way TCP/IP is set up all the current connections will be terminated when the IP address of a node changes. The creation of mobile IP [Per02] proved a clever solution to this problem which has become more apparent in recent years, with the rise of WLANs and the use of VoIP. There are a number of working groups within the IRTF (MobOpts [MO]) and the IETF (Mip4 [M4], Mipshop [MS], Mip6 [M6]) that conduct research in the field of mobile IP. Mobile IP is also part of the research conducted by the IEEE 802.21 working group [IEE06]. They do research into handoff and interoperability between heterogeneous networks. In mobile IP a node has a home agent which forwards all the packets to the nodes current IP address. Each time the node gets a new address it registers this with the home agent. The problem, however, has been the handoffs in mobile IP. Because of the time it takes to handle this handoff an overall performance drop of the TCP connection, and even packet loss, will occur [HS03]. Over the years several solutions to this problem have been suggested, such as the ones discussed by [HZS03], [YIHK02] and [Cas00]. This paper provides not only an overview but also a comparison of the solutions suggested in the literature. The remainder of this paper will be organized as follows: In section 2 a brief explanation will be given of what mobile IP is, and why it is needed. The 3rd section will explain the phenomenon of handoffs and the specifically handoffs in mobile IP. After that section 4 will list and explain briefly all

2.1 Internet Protocol


In the conventional routing mechanisms that the internet uses the IP address identifies not just the host itself, but also the subnet the host is located in. Routers determine the direction in which packets need to be sent according to part of the IP address. This concept works well for hosts with a fixed location in a network, but if a host starts to move, its address becomes incorrect in respect of the topological location in the network. This is why mobility of hosts is not supported in the conventional routing mechanisms of the internet. Schiller [Sch03] discusses several mechanisms that provide quick solutions, but none of them m eet the requirem ents. F or instance requesting a new (topologically correct) address by using DHCP; the problem with this however is that no one knows of this new address. So all the connections would be terminated, and to continue communications new connections would have to be created.

2.2 Mobile IP
To solve the mobility problem of IP a standard was proposed [Per02], namely mobile IP. We will now briefly discuss the components of the mobile IP protocol illustrated in Figure 1. MN: Mobile Node (sometimes called Mobile Host), this is the node that changes location. HN: Home Network, the network in which the Mobile Nodes (MN) home agent is positioned. This is the network in which the permanent address of the MN is located. HA: Home Agent, which is in the router of the Home Network (HN). FN: Foreign Network, the network that the MN is currently in. FA: Foreign Agent, which is in the router of the Foreign Network (FN). CN: Corresponding Node (sometimes called Corresponding Host), the node that the MN is communicating with. CoA: Care-of-Address, the temporary IP address that the Home Agent (HA) can use to contact the MN, while it resides in the

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission. 6th Twente Student Conference on IT, Enschede, 2nd February, 2007 Copyright 2007, University of Twente, Faculty of Electrical Engineering, Mathematics and Computer Science

FN. This CoA usually points to the Foreign Agent (FA), although it might sometimes point directly to the MN.

3. 4.

(pre-)allocate resources at subnet 1 and de-allocate them at subnet 0; Update a certain network-resident state of mobile nodes(e.g. an Address Registry, a path or route) at some nodes and/or transfer this state (e.g. a context) from subnet 0 to subnet 1; Ensure that all functional and non-functional properties of the service are guaranteed during and after the migration.

CN

5.

3.2 Handoffs in mobile IP


Whenever an MN moves to a new FN it knows nothing about the new network or the new router (FA). There are two methods of solving this, agent advertisement and agent solicitation. In both methods the MN eventually receives a new CoA from the new FA. Now that the MN has a new CoA it has to register the new address with the HA. If the location of the CoA is at the MN the new address will be registered directly with the HA. If the CoA is at the FA the registration with the HA will go via the FA. To register a registration request message is sent by the MN to the HA, the HA then replies with a registration reply message. T he process of the netw ork dealing w ith the m ovem ent of the MN to a new FN is called a handoff. Once the FN has received the registration reply message the handoff is complete. The handoff mechanisms will be discussed in Section 4.

Internet

FA MN CoA

HA

HN FN

3.2.1 Problems
Figure 1. Mobile IP diagram Traffic between the MN and the CN flows as indicated by the arrows in Figure 1. T he M N sends the packets directly to the CN through the FA, but the packets originating from the CN are sent to the HA and then forwarded to the MN through the FA. This type of communication is called triangular routing. Whenever an MN moves between networks a handoff must occur. This basically means that the protocol must handle the moving of the MN. Handoffs will be discussed further in Section 3. To support the mobility of nodes mobile IP has implemented a handoff mechanism. However it became apparent that this handoff mechanism was far from perfect, and could often become a bottleneck in the performance of the overall protocol, especially in a situation where handoffs frequently occur.

4. MECHANISMS FOR HANDOFFS IN MOBILE IP


The purpose of most of the solutions is to reduce the total handoff latency. This latency is made up of several factors. The total latency is the sum of the IP capability latency and the binding update latency. The IP capability latency is the time it takes before an MN can send messages from its new CoA. This IP capability latency in turn consists of the movement detection latency, the time it takes to detect that the MN has moved to another subnet, and the CoA configuration latency, the time it takes to configure a new CoA. The aforementioned binding update latency represents the time it takes to send the binding update message to the HA and to receive the confirmation message. Many of the suggestions for improvements that have been made are actually improvements on former suggestions. This is one of the reasons that a lot of the suggestions can be subcategorized. In this section all the mechanisms, and the suggested improvements to them, will be discussed.

3. HANDOFFS
As briefly mentioned in the previous section an important part of mobile IP is the handoff mechanism. We will now give a short explanation of handoffs in general in section 3.1, followed in section 3.2 by a more specific type of handoff, namely mobile IP handoffs.

3.1 Handoffs in general


Handoffs are a phenomenon seen in multiple fields of telecommunications, for instance the GSM cellular network. In GSM handoffs are needed when a mobile phone moves out of the range of one antenna and has to transfer to the service of another antenna. As mentioned in [EN02] a handoff, where a mobile node moves from subnet 0 to subnet 1, generally consists of the following five parts: 1. 2. Authenticate mobile nodes identity and check its authorization to be served by subnet 1; Register mobile node at subnet 1 and de-register at subnet 0;

4.1 Hierarchical
One improvement to the original mobile IP protocol is a change in the architecture. Compared to the original architecture the new architecture is set up in a hierarchical fashion. This improvement is mentioned and explained in both [PW99] and [TLP99]. Figure 2 is a schematic overview of the hierarchical set. As can be seen when compared to original architecture of mobile IP (represented in Figure 1) some new objects are introduced. The

most important new object is the Domain Foreign Agent (DFA) which is connected to the router of the foreign domain (FD). In mobile IP (Figure 1) each sub network has an FA connected to its router. This is not the case in hierarchical mobile IP, where the routers (R1 and R2) that are connected to the sub networks (FSN1 and FSN2) no longer have FAs. The FAs have been replaced by the DFA. Another change that has been made is the location of the CoA, which is now at the DFA. Whenever an MN moves from one sub network (FSN1) to another (FSN2) in the same domain the CoA, that the HA uses to forward messages, does not change. Because the CoA does not change the MN does not have to send a binding update to the HA, which reduces the overall latency of the handoff.
HA

4.1.1 HMIPv6
In [Cas00] and [SCEB05] a hierarchical adaptation of the mobile IPv6 protocol is discussed. In HMIPv6 a new network entity is introduced: the Mobility Anchor Point (MAP). The MAP can be considered a local HA. Binding updates (BU) are the messages that a MN sends to notify another node (e.g. an HA or CN) of its changed location. By introducing the MAP to MIPv6 the number of BUs that the MN has to send is reduced to one. The MN sends a BU to the MAP which will then redirect all traffic to the MNs new location. The MAP acts similarly to the DFA in HMIPv4 in that receives packets, encapsulates them and forwards them to the MN. HMIPv6 supports Fast Mobile IPv6 Handovers (also discussed in [SCEB05]), which allows the intra-domain handoffs between the access routers to take place smoothly. As an addition to the route optimization, which is used by MIPv6, HMIPv6 enables the MN to keep its location private while using route optimization. HMIPv6 is independent of the type of access network used and it will work transparently, as it is an extension to MIPv6. This means that if a network is not HMIPv6-capable the MN will simply use MIPv6, and vice versa.

CN

Internet HN

DFA

CoA

4.2 Fast Handoff


A second category in to which many of the solutions fall is F ast H andoff. Here the focus is on doing a lot of the work needed for a handoff before the actual handoff takes place. There are triggers or other indications to let an MN know that a handoff is about to happen, so the MN can start doing the necessary work. This reduces the IP capability latency and binding update latency, which in turn results in a smoother handoff with less chance of packet loss occurring. An example of this is the one discussed in [KP06]. After an MN has received its CoA, and has become IP capable, it will begin to scan its neighbourhood for new access routers (ARs). When the scan results in the detection of a new AR the MN is able to resolve information about this new AR. How this takes place exactly is described in detail in [KP06]. B y using the R outer S olicitation for P roxy A dvertisem ent and P roxy R outer A dvertisem ent messages, to gather information about the neighboring ARs, the MN can formulate a new CoA while it is still connected to the previous subnet. It must be noted that this will depend on the way the network is set up with regard to the assignment of IP addresses. The MN can then use the gathered information to compose a Fast Binding Update (FBU) message when the actual handoff takes place. This FBU is the message that the MN sends to the PAR (Previous Access Router) before a handoff, while it is still connected to the PAR. After the PAR receives the FBU from the MN a tunnel will be set up between the PAR and the NAR (Next Access Router). The creation of this tunnel is done by the Handover Initiate and Handover Acknowledgements messages, defined in [KP06]. When the MN is actually connected to the NAR it will use the information it already gathered to lower the IP capability latency, for instance with the use of an anticipated IP address. The registration of the new CoA with the HA occurs the same as it would in the original specification of the mobile IP protocol.

R2

R1

MN

FSN1 FD

FSN2

Figure 2. Hierarchical set up for mobile IP The type of movement by the MN discussed above is called an intra-domain handoff. In this type of handoff the previous sub network is a member of the same domain as the next sub network. When the previous sub network is not part of the same domain as the next sub network an inter-domain handoff will take place. In an inter-domain handoff the MN will have to register its new CoA (as received from its new DFA) with its HA. This occurs in the same way as it would in the regular IP protocol. Besides a reduction of the binding update latency the hierarchical set up also reduces the amount of traffic sent over the internet. Of course both these effects only occur during intra-domain handoff (movement within a domain), because inter-domain handoffs (movement between domains) are handled similarly to handoffs in traditional mobile IP. The hierarchical set up is also (part of the solutions) mentioned in [DMAD00], [RPT+00] and [ECD+02]. The suggestions mentioned in these papers all differ in small details, but in general reflect the same ideas as the ones described above.

The suggestion discussed by El Malki in [ElM06] moves to Layer 2 (the Data Link Layer of the OSI model) to smoothen handoffs. It does this by using a number of strategies that make use of Layer 2 indications to predict or rapidly respond to an upcoming handoff. One of the strategies used is called PreRegistration where the Layer 3 (MIPv4) handoff takes place before the Layer 2 handoff is completed. If the Layer 3 handoff cannot be completed in time before the Layer 2 handoff, a different strategy will be used. This strategy is called PostRegistration, where a tunnel between the two ARs is set up after the handoff. This tunnel enables the ARs to communicate after the handoff, thus reducing the packet loss. This last method is a strategy that can also be seen in [KP06], amongst others.

HN

HA

Internet

4.2.1 MIPv6
A solution similar to the ones in [KP06] and [ElM06] is discussed in [Koo05] (and in the improvement [Koo06]), the difference being that the solutions in [Koo05] and [Koo06] are adapted for MIPv6, where [KP06] and [ElM06] talked about MIPv4. An analysis and detailed explanation of the suggestion of [Koo06] can be found in [McC05]. An improvement to [Koo06] is suggested in [CZ06] which makes use of a Pre-Binding Update (PBU) message, that is sent to the CN before the handoff. It is a message that creates a temporary binding at the CN, the PBU contains the new CoA of the MN. The MN has discovered this new CoA as described above, using Proxy Router Advertisement messages.

FA CN

MN

FN
Figure 3. Triangle routing and route optimization

4.3 Route Optimization


One improvement to mobile IP discussed in many papers (for example in [PW99]) is something called Route Optimization. In mobile IP the MN can send messages to the CN via the optimal and shortest route, this is not the case for the CN. When the CN sends a packet to the MN it has to travel via the HA of the MN. If the MN is close to its HA and the CN is far away from the MN this is not really a problem. However, if the MN and CN are close to each other and the MN is far away this creates a problem known as triangle routing. Triangle routing can cause needlessly long delays in message arriving at its destination as the message from the CN has to travel all the way to the HA and then to the MN instead of straight to the MN, which would be much shorter. The difference between the two routes can be seen in Figure 3, the long route is shown in red and the optimal route in green. It is suggested that enabling a CN to have a binding for the MNs current address will solve this triangle routing problem. There are two ways in which a CN can receive a binding update. The first is when the HA receives a packet destined for a MN that is not in its HN. The HA will then send a binding update to the CN notifying it of the current CoA of the MN. Another way is when an MN changes CoA, because a handoff has taken place, or is about to take place. The MN can then send a binding update message containing its new CoA to the CN. The CN will then update its binding for the M N s address. This structure leaves some security holes that potentially allow message replaying. Message replaying is a technique which enables someone to eavesdrop on the packets that are being sent. This is why a CN needs to identify and/or authenticate the source of the binding update message. A number of papers have been written on increasing the security and enhancing the overall performance of route optimization. These include [VA06] and [AVH06].

4.4 General
Some solutions are provided in the literature that do not really fall into any of the major categories that have been introduced above. These solutions are discussed in this section.

4.4.1 Layer 2
The solution provided in [YIHK02] moves to Layer 2 (of the OSI model) to improve handoffs without changing the existing mobile IP specification. Basically it uses a database connected to a MAC-bridge which keeps track of all the MAC-addresses currently involved in a handoff. When it receives a MAC-frame that is directed towards a MAC-address that is currently in the database it directs the frame to the correct port.

4.4.2 Buffers
An improvement that is often part of the solutions suggested is the use of buffers located at the previous FAs during handoffs. These are used to catch packets destined for the MN. When the MN is connected to a new FA, and the previous FA is notified of the new address, the previous FA will forward the buffered packets to the MN. A different use of buffers is suggested in [TLP99]; here an MN is assigned a multicast address by the DFA. The router to which the MN is connected informs its neighbours of this multicast address. When the DFA receives a new message destined for the MN it sends this message to the M N s multicast address. The MN receives the message from its router, however the neighbours of this router also receive this message (because they also listen for this multicast address) and they buffer the message. Now when the MN moves to a neighbouring network/router this router has buffered the last few messages that were destined for the MN. This can drastically reduce packet loss during handoff.

A number of papers have suggested the use of Route Optimization in combination with buffers to reduce packet loss, these papers include [ELS+01] and [WCLY02].

this scheme can interact with other types of networks with the primitives proposed by 802.21 [IEE06]. Another paper that discusses adaptation of the mobile IP protocol for a different type of network is [YD06], in which 3G CDMA is the chosen network. As in [JJH+06] [YD06] also talks about the changes needed for fast handoff to work in 3G CDMA. In addition to the fast handoff method another strategy, S elective bi-casting, is also discussed. S elective bi-casting basically involves a form of buffering w hich w ill reduce the amount of packet loss during a handoff.

4.4.3 Other solutions


A solution that is a combination of a hierarchical set up and a fast handoff strategy is discussed in [HZS03]. In addition to the combination of the hierarchical set up and the fast handoff strategy a new movement detection algorithm is also introduced. The authors of [HZS03] use simulation and mathematical analysis to show that this combined scheme can provide lossless handoffs. In [SZC04] the original mobile IP specification is tweaked and adapted in relatively small ways to improve the overall performance of the protocol with regard to packet loss, speed drop and other properties. One of the things that is improved is the movement detection latency. It does this by circumventing the restrictions for solicitations and advertisements as opposed by the mobile IP protocol. This solution only applies for infrastructure-mode wireless LANs, which means it is a solution for micro-mobility only. Micro-mobility is mobility inside a domain, with access point. As opposed to macromobility, which is what a hierarchical set up can provide. HAWAII is the name of the scheme suggested in [RPT+00], it stands for Handoff-Aware Wireless Access Internet Infrastructure. It greatly resembles the hierarchical set up that has been explained before, with only a slight difference in the way packets are routed within a domain. In addition to the hierarchical set up improvement HAWAII also implements the Route Optimization strategy. Finally in [QZDK06] a solution is described for a problem that has not really been discussed here. This problem is not directly a consequence of the handoff protocol in mobile IP, but more a consequence of some of the suggested improvements to the handoff protocol. It talks about M IP 6 location privacy solutions. It provides som e practical solutions to prevent som eone from easily tracking a nodes m ovem ents w hen that node uses MIPv6. A specific situation that is sensitive to the invasion of privacy is in Route Optimization when the MN or HA notifies the CN of the new CoA of the MN. By listening in to this message a third party could determine the location and movement of an MN. Another privacy-sensitive piece of knowledge is the IP address of the HA of an MN, this is also something that should not be revealed to a third party in view of privacy. A number of these situations are discussed in [QZDK06] and solutions are provided.

5. CRITERIA FOR HANDOFF MECHANISMS


To make a proper comparison six criteria were set up that determine how good a mechanism is. We will briefly explain each of these criteria and why they are factors to take in consideration when making this comparison.

5.1 Implementation
This criterion determines how easy it is to implement the protocol into current hardware and software. Some protocol adaptations need only to change certain parts of the architecture, like just the software on the mobile host, while others introduce entirely new objects into the network architecture. The amount of effort needed to implement the protocol determines this criterion, less is better.

5.2 Standards adaptation


Because standards have already been determined in the field of mobile IP, an important factor to consider is how much the existing standards need to be adapted to support the new or updated protocols. In an ideal situation the currently existing protocols would not have to be adapted at all.

5.3 Scalability
This criterion reflects how well the protocol can handle a major increase in usage. A protocol with good scalability can handle a large amount of hosts using the protocol at the same time, without severely affecting the overall performance. An important factor here is the amount of traffic generated by the protocol when handling a handoff. The less traffic generated the better the scalability of the protocol. Another factor that could limit the scalability is the need for a central database of some sort, because this database and the agent handling the requests for this database have a fixed capability.

5.4 Handoff latency


This is probably the most logical criterion, as this is the objective of most of the new or updated protocol suggestions. Decreasing the handoff latency generally improves the smoothness of the experience for the user. Of course there are other weighing factors when determining this smoothness, such as the amount of packet loss. For instance, a protocol with a 50 percent lower latency but 100 percent packet loss during the handoff will not be a good solution in most situations.

4.5 Other networks


As mentioned before much of the theory and experience with handoffs in mobile IP came from GSM and cellular networks in general, which is where the origins of handoffs lie. This is why we will also take a brief look at how handoffs are implemented in other types of networks than the typical WLANs. One of the suggestions made to improve the handoff latency in MIPv6 was the Fast handoff method. This strategy makes use of Layer 2 to predict and prepare for future handoffs. However the specific link-layer information available depends on the type of network. In [JJH+06] an adaptation of the fast handoff mechanism for 802.16e networks is discussed. The protocol that is officially called 802.16e is also know n as M obile W iM A X . The authors discuss the changes needed for fast handoff to work in 802.16e networks. They also explain how

5.5 Packet loss


Another important objective for many of the suggested adaptations or new protocols is to reduce the amount of packet loss that occurs during a handoff. Packet loss can occur during or after the actual movement of the mobile host.

5.6 Packet reordering


A criterion that does not affect some protocols at all but can affect others quite a lot is the possibility of packet reordering. Packet reordering can occur when the host is connected to a new subnet and receives packets that are forwarded from the old subnet. In most situations this criterion is closely related to the aforementioned criterion, the occurrence of packet loss.

This criterion is closely related to the previous criterion, because they concern the necessary amount of change to the current situation. The difference however is that this criterion focuses on the protocols transparency to older objects in the network architecture. The hierarchical solution is a good one because the only thing that is really different from the original mobile IP specification is the CoA registration process and the location of the actual CoA. The registration of the CoA is a process that does not really involve the FAs, so no adaptation to the current FAs is required. This means the hierarchical set up is transparent to the FAs. The FAs are actually just routers in the hierarchical set up, as can be seen in Figure 2. The fast handoff method does require an FA to know a number of extra messages that have been introduced for fast handoffs. However the protocol can handle handoffs between two FAs where only one is capable of the fast handoff protocol. So in this respect it is good in that it can work with older network objects. Route optimization is a good suggestion because it is transparent to the FAs. However the CN does need to be capable of handling the binding update messages, because the HA will reply with a binding update whenever the CN tries to send a packet to the MN. The HA will not forward the message to the MN, so if the CN does not understand the binding updates the packets will never be delivered. In the general solutions section (4.4.1) the Layer 2 solution by [YIHK02] was introduced. One of the objectives of this solution is to not adapt the mobile IP specification. It accomplishes this by moving to the Layer 2 of the OSI-model in order to improve the handoff latency for the Layer 3 ((mobile) IP layer). In [SZC04] (discussed in section 4.4.3) a strategy is introduced that does not require a change in the mobile IP specification. The combination of this fact and the results of simulations discussed by the authors make this a good solution with regard to this criterion. The solutions that make use of buffers (section 4.4.2) are generally transparent to FAs that do not use buffers. The solution suggested in [HZS03] needs quite a lot of adaptation of the standard as this is a combination of different mechanisms that all need their own changes in the standards. The solution discussed in [RPT+00] can be compared to the hierarchical set up with regard to the amount of adaptation of the standards that is needed and the protocols transparency. Overall the solution mentioned in [SZC04] is a good solution that does not require any change in the standards while performing well in simulations. However this solution only provides micro-mobility. For macro-mobility the hierarchical set up is a good solution with regards to the required amount of adaptation to the current standards.

6. COMPARISON OF MECHANISMS
In the previous section the six criteria for this comparison have been explained. This section will report on the actual comparison of the mechanisms discussed in section 4 based on the criteria discussed in section 5.

6.1 Implementation
How easy a protocol is to implement has a lot to do with the number of objects that need to be added or changed in the architecture with regard to the original protocol. In this definition the hierarchical solution would be a good one. The reason for this is the fact that in a domain only the DFA is added. The routers or access points do not have to be updated or changed in this situation. The fast handoff is less ideal with regard to the implementation of the protocol. The FAs in the network need to be altered so they can deal with the new types of messages that the protocol introduces, for instance the Proxy Router Advertisement messages. These are the messages that can tell an MN crucial information about the FA. This information is needed to achieve a smooth handoff as defined by the fast handoff protocol. The fast handoff method with use of layer 2 also requires an adaptation of the FAs. So with regard to the implementation it is not much better than the original fast handoff suggestion. Route optimization does not require any changes in the FAs, only the MN, HA and CNs. However any changes required in the CNs are detrimental to the ease of implementation. Because this would mean that only updated hosts could communicate with hosts that use mobile IP. This is not a very good solution, therefore, with regard to the ease of implementation. The solutions that involve the use of buffers, which are discussed in section 4.4.2, are generally not very good. The reason for this being that they all require some form of adaption in the FAs, because this is where the buffers would have to be implemented. As for the miscellaneous solutions described in section 4.4.3 ([HZS03], [SZC04] and [RPT+00]) they all require a relatively large amount of change needed to several objects in the network architecture. Overall, the hierarchical set up would be the most implementable solution because it requires the least amount of change to the current situation.

6.2 Standards adaptation


Another question that has to be asked is how much the standards would need to be adapted. This is important because routers already exist today that have implemented the existing standards and protocols. A solution that would render these existing routers useless would not be a good solution in this respect. A solution that can deal w ith older rou ters, that only know the original mobile IP standard, would be a much better solution.

6.3 Scalability
In general, solutions without a central database or server perform better when it comes to scalability. The amount of traffic generated is also a factor when it comes to the scalability of a certain protocol. In one way the hierarchical set up is a good solution because it reduces the amount of traffic sent over the internet. This is caused by the fact that intra-domain handoffs do not generate any traffic outside the domain; there is no need to register with

the HA. However there is also a bad side to this solution because it involves the use of a DFA which has to handle all the MNs within a domain. This DFA has a certain capacity with regard to the amount of MNs and binding updates it can handle. Of course the FAs in the original mobile IP scheme also have a limit to their capacity but they are lower in the architecture, so they have to handle a smaller number of MNs. The fast handoff method can be considered a good solution with regard to scalability because it does not involve a central database or server, only the FAs. On the other hand in a situation where there are a lot of handoffs the fast handoff method is not so good, because of all the traffic generated for the binding updates with the HAs of the MNs. Route optimization and the solutions mentioned in 4.4.2 and 4.4.3 have no major issues with scalability because there is no central point or database that can become a bottleneck. An exception here is the solution mentioned in [WCLY02]. It introduces an MRT which can be described as a database system. However as these databases are located at the edges of the network (in the FAs and HAs) they do not influence the scalability as much as a central database would. Another solution that contains databases is the Layer 2 solution mentioned in [YIHK02] (and mentioned in section 4.4.1). However the databases will not affect the scalability in a large way here because they only contain the MAC-addresses of the MNs currently involved in a handoff. This will only form a problem in a situation where a lot of handoffs take place at once. In a situation with a lot of movement and small sub-networks a hierarchical set up would be better. However in a situation with large sub-networks and a large amount of MNs the fast handoff solution would be better. Overall, the solution described in [HZS03] would probably be the best here, as this combines the different strategies that were discussed above.

both would be the best option, so the solution introduced in [HZS03] seems to be the right choice.

6.5 Packet loss


In general, solutions that use buffers can prevent packet loss. The hierarchical set up approach does not really help reduce packet loss except by reducing the CoA registration latency. This in turn reduces the amount of time that packets cannot arrive at the correct address, thus reducing the packet loss. Fast handoff does help reduce packet loss by decreasing the IP capability latency and also by implementing the use of tunnels between FAs to forward packets destined for the MN. If buffers are not used the route optimization might actually increase the chances of packet loss. This is because the binding update message needs to be sent and processed before the CN will send packets to the correct new address of the MN. Packets sent before the binding update message arrives at the CN will be received by the previous FA. However, many solutions use buffers or tunnels to enable the FAs to forward the packets to the MN. In [HZS03] the authors say that their protocol can provide completely lossless handoffs. So in general the protocols that use buffers emerge as good contenders here, and with the [HZS03] suggestion claiming lossless handoffs it again seems like a very good solution.

6.6 Packet reordering


In the comparison based on the previous criterion it was mentioned that buffers can help with packet loss. Buffers also have some disadvantages, however, including the chance of packet reordering occurring. The hierarchical set up does not use buffers and, consequently the chances of packet reordering occurring are not increased by the hierarchical set up. The fast handoff method can cause packet reordering. This can occur when packets from the previous FA are forwarded to the next FA. The route that those packets take is then usually longer than the route of the packets that go straight to the next FA. Another factor that will cause the first packets to arrive later is that they usually wait in the buffer until the MN is ready to receive them. This causes the second packets to arrive earlier, which implies packet reordering. The same problems occur with the other solutions that use some sort of packet forwarding or buffers. With regard to packet reordering the solutions that do not use buffers, or tunneling of packets from previous FAs, are generally better.

6.4 Handoff latency


As explained in the first part of section 4 the total handoff latency is made up of multiple latencies. For instance the hierarchical set up solution reduces the binding update latency by allowing the MN to register its new CoA with an agent closer to its own location. Which is the DFA instead of its HA. Because the DFA is closer to the MN the time it takes to register the CoA will be reduced, and thereby the overall latency will be reduced. The fast handoff method addresses a different latency, namely the IP capability latency. Because the actual handoff takes place much faster the time that the MN is not capable of receiving any messages is greatly reduced, which in turn reduces the overall latency. Route optimization does not really improve the overall handoff latency. Actually the only issue here is that when you consider the time it takes for an MN to send a binding update to a CN (and for the CN to process it) as part of the overall latency, the overall latency will increase. The other solutions mentioned in section 4.4 rely on either the hierarchical set up or the fast handoff method to reduce the latencies. There are two exceptions to this which are [YIHK02] and [SZC04]; they both try to reduce the movement detection latency, each in their own way. Both the hierarchical set up and the fast handoff method have shown to reduce the overall latency. Again a combination of

7. CONCLUSIONS AND FUTURE WORK


In overview the hierarchical set up appears to be the best discussed solution when it comes to the ease of implementation. Because the solution provided in [SZC04] does not change the standard IP protocol and does increase the smoothness of handoffs significantly, it appears to be the best choice when it comes to the amount of changes that have to be made to the current standards. In respect of scalability, handoff latency and packet loss the combination of strategies that has been introduced in [HZS03] appears to be the best option. So in conclusion the hierarchical set up is a good addition to the existing protocol, which is most effective in a situation with a lot of local movement of the mobile nodes. However the solution suggested by [HZS03] can be considered the overall winner, as it includes a hierarchical set up among other proven strategies.

A conclusion that can be drawn here is that in many respects the suggested solutions need not be exclusive. A combination of strategies can be used. This also follows from the simulation results discussed in e.g. [HZS03]. This illustrates that the different strategies form a synergy that is often more successful than the implementation of a single strategy. With a view to the future it is also important that enough research and development is conducted regarding the smooth cooperation between the different types of networks (as mentioned in section 4.5).

[IEE06] IEEE 802.21 working group, http://www.ieee802.org/21/, accessed October 2006. [JJH+06] Jang, H., Jee, J., Han, Y., Park, S.D., Cha, J. M obile Ipv6 F ast H andovers for IE E E 802.16e N etw orks, Internet D raft, M IP S H O P W orking G roup, 2006. [Koo06] K oodli, R ., Fast Handovers for Mobile IPv6, Internet Draft, MIPSHOP Working Group, 2006 [K oo05] K oodli, R . F ast H andovers for M obile IP v6, R F C 4068, Network Working Group, 2005. [K P 06] K odaly, R ., P erkins, C . M obile IP v4 F ast H andovers, Internet Draft, Mobile IPv4 Working Group, 2006. [M 4] M obility for IP v4 (m ip4) C harter, http://www.ietf.org/html.charters/mip4-charter.html, accessed September 2006. M obility for IP v6 (m ip6) C harter, http://www.ietf.org/html.charters/mip6-charter.html, accessed September 2006.

ACKNOWLEDGEMENTS
We thank the Association of Computing Machinery for making available a proceedings template [ACM] that we could use for the TC&CR Conference. We also thank Georgios Karagiannis for his help and feedback as a supervisor. Finally, we would like to thank Teake Blom, Dirk Engels and Jeroen Logtenberg for their valuable feedback.

[M6]

REFERENCES
[A V H 06] A rkko, J., V ogt, C ., H addad, W . A pplying Cryptographically Generated Addresses and CreditB ased A uthorization to M obile IP v6, Internet D raft, Network Working Group, 2006. [C as00] C astelluccia, C . H M IP v6: A hierarchical m obile IP v6 proposal , ACM SIGMOBILE Mobile Computing and Communications Review, volume 4, p.48-59, 2000. [C Z 06] C hen, H ., Z hang, J., Prep-Binding of Fast Handovers for Mobile IPv6, Internet D raft, M IP S H O P W orking Group, 2006. [DMAD00] Das, S., Misra, A., Agrawal, P., Das, S.K. T eleM IP : T elecom m unications-Enhanced Mobile IP A rchitecture for F ast Intradom ain M obility, IEEE Personal Communications, August 2000, p.50-58, 2000 [ECD+02] Ergen, M., Coleri, S., Dundar, B., Puri, A., W alrand, J., V araiya, P . P osition leverage smooth handover algorithm for m obile IP , IEEE ICN, 2002. [E lM 06] E l M alki, K . Low Latency Handoffs in Mobile IPv4, Internet D raft, N etw ork W orking G roup, 2006 [ELS+01] Eom, D., Lee, H., Sugano, M., Murata, M., M iyahara, H . Im proving T C P handoff perform ance in M obile IP based netw orks, Computer Communications, volume 25, issue 7, p.635-646, 2001 [EN02] E ndler, M ., N agam uta, V . G eneral approaches for im plem enting seam less handover Proceedings of the second ACM international workshop on Principles of mobile com puting , p.17-24, 2002 [HS03] Hsieh, R. and Seneviratne, A . A C om parison of Mechanisms for Improving Mobile IP Handoff Latency for End-to-E nd T C P , Proceedings of the 9th annual international conference on Mobile computing and networking, p.29-41, 2003. [HZS03] Hsieh, R., Guang Zhouand, Z. and Seneviratne, A. S -MIP: A Seamless Handoff Architecture for Mobile IP , INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and Communications Societies. IEEE, volume 3, p.17741784, 2003.

[M cC 05] M cC ann, P . M obile IP v6 F ast H andovers for 802.11 N etw orks, R F C 4260, N etw ork W orking G roup, 2005. [M O ] IP M obility O ptim izations (M ob O pts) R esearch G roup, http://www.irtf.org/charter?gtype=rg&group=mobopts, accessed September 2006. [MS] M obility for IP : P erform ance, S ignalling and H andoff O ptim ization (m ipshop) C harter, http://www.ietf.org/html.charters/mipshopcharter.html, accessed September 2006. [P er02] C . P erkins, IP Mobility support for IPv4, RFC 3344, Network Working Group, 2002 [PW99] Perkins, C., Wang, K. "Optimized Smooth Handoffs in Mobile IP", The Fourth IEEE Symposium on Computers and Communications, p. 340, 1999. [Q Z D K 06] Q iu, Y ., Z hao, F ., D avis, U ., K oodli, R . M obile IP v6 L ocation P rivacy S olutions, Intern et Draft, MobOpts Working Group, 2006. [R P T 00] R am jee, R ., P orta, T . L a, T huel, S . IP M obility support using H A W A II, Internet D raft, IE T F M obile IP, 2000. [SCEB05] Soliman, H., Castelluccia, C., El Malki, K., Bellier, L . Hierarchical Mobile IPv6 Mobility Management, RFC 4140, Network Working Group, 2005. [S ch03] S chiller, J. C hapter 8.1 M obile IP in M obile Communications, 2nd edition, P earson E ducation Limited, Harlow, England, 2003. [SZC04] S harm a, S ., Z hu, N ., C hiueh, T . L ow -latency mobile IP handoff for infrastructure-m ode w ireless L A N s, Selected areas in Communications, IEEE Journal on, volume 22, p.643-652, 2004 [TLP99] Tan, C.L., Lye, K.M., Pink, S. "A fast handoff scheme for w ireless netw orks, Proceedings of the 2nd ACM International workshop on Wireless mobile multimedia, p.83-90, 1999 [V A 06] V ogt, C ., A rkko, J. A T axonom y and A nalysis of E nhancem ents to M obile IP v6 R oute O ptim ization, Internet Draft, Network Working Group, 2006.

[WCLY02] W u, I., C hen, W ., L iao, H ., Y oung, F . A seamless handoff approach of Mobile IP protocol for mobile w ireless data netw orks, Consumer Electronics, IEEE Transactions on, volume 48, p.335-344, 2002 [Y D 06] Y okota, H ., D om m ety, G . Mobile IPv6 Fast Handovers for 3G CDMA Networks, Internet D raft, Network Working Group, 2006.

[YIHK02] Yokota, H., Idoue, A., Hasegawa, T. and Kato, T. L ink L ayer A ssisted M obile IP F ast H andoff M ethod over W ireless L A N N etw orks, Proceedings of the 8th annual international conference on Mobile computing and networking, p.131-139, 2002.

Das könnte Ihnen auch gefallen