(NetLMM) working group of the Internet Engineering Task ... the host based mobility management protocols in use today, ... traffic volume and no tunneled packets in the access network. .... Air interface traffic overhead refers all control data.
Evaluation of a Network based Mobility Management Protocol : PMIPv6 Asanga Udugama, Muhammad Umer Iqbal, Umar Toseef, Carmelita Goerg Communication Networks Group – University of Bremen Bremen, Germany [adu | mui | umr | cg ]@comnets.uni-bremen.de Abstract—The Proxy Mobile IPv6 (PMIPv6) is a network based mobility management protocol standard that was ratified recently by the Network-based Localized Mobility Management (NetLMM) working group of the Internet Engineering Task Force (IETF). PMIPv6 is a protocol that uses the same concepts as used in Mobile IPv6 (MIPv6), but modified to operate in the network part only instead of involving the Mobile Node (MN) as well. PMPv6 is claimed to posses a number of advantages over the host based mobility management protocols in use today, above all MIPv6. The main advantage of using PMIPv6 is the freeing up of the mobile host in doing any mobility related activities and thereby saving its resources. The saving of resources may result in their usage for other purposes or even enable otherwise capabilities restricted devices to operate in the PMIPv6 domains. Other advantages include reduced signaling traffic volume and no tunneled packets in the access network. These aspects become very important since the access networks in mobile networks usually are air interfaces. Further, PMIPv6 is also becoming a very attractive mobility management protocol for mobile network operators as seen by its inclusion in current 3rd Generation Partnership Project (3GPP) standardization as a possible alternative mobility management protocol for the Long Term Evolution (LTE) technologies. In addition to qualitative analyses and comparisons, the work attempts to quantify these advantages to show the achieved benefits. The quantifications are done through measurements in a real test-bed which is installed with a PMIPv6 implementation developed as part of this work. Keywords - Network based mobility management; PMIPv6; NetLMM; Implementation; Evaluation
I.
INTRODUCTION
The tremendous advancements in the field of communication and information technology over the last decades have influenced our lives greatly. Once, the concept of communication was limited to fixed line analogue connections with low data rates and poor transmission quality with the only objective of transmitting voice between the two ends. Today, the scenario has completely changed. High speed connections are available, where transmission of all kinds of multimedia traffic is possible even over wireless interfaces with a touch of a button. The advancements in the field of microelectronics have also provided users with cost effective handheld devices capable of handling multimedia traffic while maintaining mobility. These technological advancements have brought about many new opportunities for users as well as the providers of communication services. At the same time, new
Changpeng Fan, Morten Schlaeger Nokia Siemens Networks Berlin, Germany [changpeng.fan | morten.schlaeger ]@nsn.com challenges such as increased usage of services by users and providing a satisfactory quality of service for all the users need to be addressed. It is becoming obvious that all future technologies and services related to the field of communication and information technology will, to a large extent center around wireless based networks and paradigms. These new technologies will pose new challenges for the service providers. Some of the major issues that will affect selection of those technologies are the available resources in the user’s devices and the cost of bandwidth. Over the last decade “mobility management” has become an important area of research and various new protocols and standards have been developed. Different network architectures have also been proposed to maximize the efficiency of mobility management techniques. Generally, mobility management refers to the techniques through which users of a particular service can move around without noticing any change or disconnection in the service. Mobility management can be either network based or terminal based. In the former only network is involved in the mobility management signaling whereas in the later the terminal also participates. The importance of network based mobility management over terminal based can be understood from the scenario that follows. Consider an Internet Protocol version 6 (IPv6) network using Mobile IPv6 (MIPv6) as its mobility management protocol [1]. Suppose a MIPv6 capable node connects to the network over a wireless interface. The node, after configuring its address sends a location registration message to its home agent. The home agent after processing the location registration sends an acknowledgement message to the mobile node. As long as the Mobile Node (MN) stays connected to the link, it sends and receives periodic location update messages and acknowledgements. Whenever the MN moves and connects to some other part of network, it exchanges these same messages. Furthermore, all the traffic to and from the MN is tunneled traffic, when using basic functionality of MIPv6. MIPv6 is a terminal based mobility management protocol and the following drawbacks can be noted from above. • Periodic exchange of location update messages and tunneled traffic over the air interface reduces available bandwidth
This work is partly sponsored by the German Ministry for Education and Research (Bundesministerium für Bildung und Forschung, BMBF) within the framework ‘Network of Tomorrow’.
978-1-4244-2517-4/09/$20.00 ©2009 IEEE
• •
MNs are usually resource restricted and their involvement in mobility management process consumes battery and incurs processing overheads MNs must have additional mobility managements support in their protocol stacks (e.g. MIPv6 functionality)
The IETF recently standardized a network based mobility management protocol called the Proxy Mobile IPv6 (PMIPv6) [2] through the Network-based Localized Mobility Management (NetLMM) work group. PMIPv6 reduces traffic in the access networks as the MN does not perform any signaling and also the increase of available bandwidth due to reduced signaling traffic and no tunneled packets in the access network. Since most of the modern mobile devices have restricted resources such as memory and processing power, not performing mobility management on their own is very attractive. This protocol has further come to the limelight due to its attractiveness for mobile network operators. Through the 3GPP standardization process, PMIPv6 is being promoted as a possible alternative for newer 3GPP and non-3GPP wireless technologies. The main objective of this work is to evaluate the performance of PMIPv6 [2] against the host based mobility management protocol, MIPv6 [1]. In this work, a PMIPv6 implementation based on the RFC 5213, which is one of the firsts or possibly the first of its kind, has been developed and evaluated against an existing MIPv6 implementation to quantify some of the benefits associated with PMIPv6. This paper is structured as follows. After the introduction, Section II looks at some of the mainly network based mobility protocols in existence today with a comparison of their features. Section III then gives a brief description of the PMIPv6 protocol. In Section IV, the details of the PMIPv6 implementation and the test-bed are presented. Section V follows by describing the performance results as well as their analyses. Section VI concludes with a short summary. II.
RELATED MOBILITY MANAGEMENT PROTOCOLS
PMIPv6 is considered as a very attractive network based mobility management protocol by 3GPP and IETF. This is mainly due to its ease of deployability in current IPv6 capable user devices without the need for any changes. This section looks at some of the currently discussed related protocols, especially focusing on network based mobility management protocols. • MIPv6 [1]: This is a host based mobility protocol which is intended for IPv6 capable nodes. MIPv6, which was the basis for PMIPv6, is still mostly in the research domain without any real deployment. • IP2MM [3]: This is a network based mobility management protocol designed for next generation mobile network architectures. This protocol intends to provide mobility management for a proposed network architecture known as IP2, which is an IP based network platform for future mobile networks. • Fast Handover for MobileIPv6 [4]: This is a network assisted mobility management protocol and help MNs to
•
acquire new IP addresses before performing handovers. Thus using this protocol allows MNs to have fewer packet losses. MobiSplit [5]: This is a mobility management architecture, which suggests mobility management to be done in a modular approach.
Table 1 compares the features of above protocols in brief. This comparison is made to analyze the usability for PMIPv6 and also to incorporate new ideas and functionalities in it which are present in other protocols. A global mobility management domain refers to a network domain where the mobility management techniques have to resolve inter access network movements. A local mobility management domain refers to a network domain where the movement of the MN is restricted within one access network only. Multi-homing enables a MN to utilize multiple addresses to attach to the networks. Air interface traffic overhead refers all control data that are sent via the wireless interface. Tunneling overhead here refers to the reduction in usable throughput and the processing power needed by a MN to process tunneled packets. As mentioned, a MN is a resource restricted entity and processing of tunneled packets consumes both, the processing power and the battery. Therefore this parameter is considered in the analysis of this work. Table 1 Comparison of mobility management Protocol Features
MIPv6
MobiSplit
IP2MM
FMIPv6
PMIPv6
Local/Global domain Packet loss in Handover
global
both
global
global
local
Yes
Yes
Yes
No/Yes
Yes
Multi-homing
No
Yes
No
No
Yes
Air interface Traffic overhead Terminal modifications Tunneling overhead at MN
High
Low
Low
High
Low
Yes
No
Yes
Yes
No
High
Low
No
High
No
The usability of PMIPv6 is evident from the Table 1 and makes PMIPv6 a suitable mobility management protocol for the future networks. III.
PROXY MIPV6 PROTOCOL
The PMIPv6 protocol [2] was recently accepted as a standard by the IETF. The operations of PMIPv6 are similar to that of MIPv6 and in fact, reuse the functionality described in the MIPv6 standard [1]. Compared to MIPv6, PMIPv6 is designed to bring about the following advantages. • No necessity to have mobility functionality in the protocol stack of a MN • Considerable reduction in the signaling overheads over the access network to which MNs get connected which are usually air interfaces • Reduced requirement for processing and resource overheads at the MN The PMIPv6 protocol introduces a number of new concepts and mechanisms that are different from MIPv6. A brief description of these is provided in the following.
•
A Mobile Node (MN) is an IPv6 node and can be any handheld device such as a laptop computer, a personal digital assistant or a mobile phone. A MN is capable of communicating over wireless interfaces and can change its position while communicating. The MN is not expected to take part in any mobility related signaling.
• • • •
Policy Profile is a database maintained by network which contains information about different parameters of the MN, required by MAG and LMA to provide services to it. Proxy Binding Update (PBU) is a binding update sent by a MAG to a LMA on behalf of a MN. Proxy Binding Acknowledgement (PBA) is a binding acknowledgement sent by a LMA to a MAG in response to PBU. Per-MN-Prefix and Shard-Prefix-Model refers to the prefixes assigned to MNs are scoped. With a per-MNprefix model, a MN-HNP is assigned to one MN only, where as in a shared-prefix-model, the MN-HNP is shared between different MNs.
The Figure 1. shows these concepts of a PMIPv6 domain, graphically. IV.
IMPLEMENTATION
A PMIPv6 implementation was developed to evaluate the performance of the protocol as part of this work. This implementation was based on [2]. Further, it was developed by extending an existing MIPv6 implementation. The extensions that were done to the MIPv6 implementation are explained briefly in the following. Figure 1. Concepts of PMIPv6
•
•
•
•
•
•
•
Local Mobility Anchor (LMA) in PMIPv6 is same as a Home Agent (HA) in MIPv6 with some extended functionalities. The LMA hosts MN’s prefix, maintain binding association of the MN and forward traffic to and from the MN. Mobile Access Gateway (MAG) is a network entity that is responsible for mobility management on behalf of the MN. All signaling needed for the location update of MN is communicated to LMA by MAG. A MN is connected to a PMIPv6 domain via the MAG. Local Mobility Anchor Address (LMAA) is the address configured at the LMA, which is used by the MAG to send mobility related signaling on behalf of the MN. LMAA is also the end point of the bi-directional tunnel established between LMA and MAG. Proxy Care of Address (PCoA) is the address configured at the MAG. It is also the end point of the bidirectional tunnel between the LMA and the MAG. The PCoA is considered by the LMA as a Care of Address (CoA) of the MN. Home Address of Mobile Node (MN-HoA) is the address used by a MN for its communications. The MN can use this address at all attachment points in the PMIPv6 domain. Home Network Prefix (MN-HNP) is a prefix advertised by every MAG in a PMIPv6 domain for a MN. The prefix is topologically anchored at LMA and a MN will always configure its address from this prefix. Mobile Node Identifier (MN-NAI) is an identifier used by a MN to perform the authentication procedures with the PMIPv6 network entities such as LMA and MAG.
A. Implementation of LMA functionalities The MIPv6 HA functionality was modified for the MIPv6 HA to act as a LMA. Following is the list of these changes. 1. Addition of new mobility options 2. Addition of new fields in binding cache 3. Addition of new functionalities to handle the time stamp option 4. Modifications to handle PBA messages instead of MIPv6 BA messages (structure and functionality) 5. Modifications to handle PBU messages instead of MIPv6 BU messages (structure and functionality) B. Implementation of MAG functionalities The MIPv6 MN functionality was modified for the MIPv6 MN to act as a MAG. Following is the list of these changes. 1. Addition of new mobility options 2. Addition of new fields in binding update list 3. Addition of functionalities for detecting MN movements 4. Modifications to handle PBU messages instead of MIPv6 BU messages (structure and functionality) 5. Modifications to handle PBA messages instead of MIPv6 BA messages (structure and functionality) C. Implementation of MN functionalities In PMIPv6, MN is not directly involved in any mobility related signaling exchanges with its mobility anchor (LMA). However, a mechanism is needed so that the MN can register itself with a MAG using its MN-NAI. The registration of the MN-NAI is important for validating the MN. Moreover, a mechanism is also needed by MAG to detect the presence or absence of a MN in the PMIPv6 domain. Since, how to achieve these two requirements are not addressed in [2], two approaches are devised to detect MN connectivity.
Approach 1: A software entity called the User Enrollment Manager (UEM) is run at the MAG. In this approach, the UEM waits for a trigger from the MN. This trigger message is a special ICMPv6 message called the NAI-Advertisement (NAI-Adv) that contains two options. One option carries the MN-NAI and the second option contains a request for joining or leaving the PMIPv6 domain managed by the MAG. When a MN wants to connect to a MAG, it sends a NAI-Adv message that carries its MN-NAI with the join option set. On receipt of this NAI-Adv, the UEM will forward the information to the MAG. The MAG validates the MN-NAI and based on the request for joining, proceeds with further steps i.e. creating binding update list, sending PBU, etc. When the MN wants to leave the PMIPv6 domain, it sends another NAI-Adv with the leave option set. UEM passes this request to the MAG and MAG immediately sends a de-registration request to the respective LMA. In this approach, the MN has to explicitly send join or leave request to MAG each time. Table 2 Test-bed setup (hardware 6 software used) Hardware Configurations
LMA / MAG / HA
MN
Processor
AMD Athalon 64 processor
Intel Centrino 1.8 GHz
RAM
1 GB
512 MB
WLAN
-
802.11b on 11Mbps
-
HSDPA/UMTS/EDGE Datacard
HSDPA Software Configurations
LMA / MAG / HA
MN
Operating System
Debian Etch with Linux 2.6.23 kernel
Debian Etch with Linux 2.6.23 kernel
PMIPv6 / MIPv6
pmip-0.1 / nemo-0.2
nemo-0.2
Packet Capture
wireshark 0.99.4
wireshark 0.99.4
Packet Generator
iperf 2.0.2
iperf 2.0.2
IPv6-over-IPv4 for Commercial HSDPA
openvpn 2.0.9
openvpn 2.0.9
Approach 2: In this approach, as opposed to the previous one, MN does not have to explicitly send the join or leave request to UEM. The MN, in this approach continuously sends NAI-Adv messages to the UEM with a user configurable interval (e.g. 2 seconds). These messages only contain MNNAI as an ICMPv6 option. The UEM continuously listens for these messages from the MN. If it does not receive any MNAdv for a specified period (e.g. 4 seconds), it concludes that the MN has moved out of the PMIPv6 domain. The UEM informs the MAG, which in turn sends a de-registration request to LMA to end the session. D. Test-bed setup The performance of PMIPv6 is evaluated against the performance of MIPv6, in a real test-bed. This test-bed is installed with PMIPv6 and MIPv6 as shown in Fig. 2. Table 2 shows the hardware and software configurations of the testbed. The MN is configured to use WLAN and HSDPA connections to connect to the PMIPv6 domain. A commercial HSDPA connection is used and hence an IPv6-over-IPv4 tunnel has to be established to carry IPv6 packets. This is done
using openvpn. A multi-purpose packet generator called iperf is used to generate different types of user traffic. A packet capture tool called wireshark is used to capture the packet flow for analysis.
Figure 2. Test-bed setup
V.
ANALYSIS OF RESULTS
NetLMM, through its problem statement identifies a number of advantages on having a network based mobility management approach compared to a terminal based approach. One of the main advantages is the improved performance on the MN side. The focus of this section is to quantify the benefits for the user of the MN. In this analysis, the performances of transport layer protocols are measured. Since MIPv6 is the most widely researched protocol, we have used MIPv6 as the protocol to compare the performance of PMIPv6. The previously described test-bed is used for the measurements. The following measurement tests are done. • • • •
UDP throughput: Checks the improvement of throughput that is gained by the applications under a given rate UDP packet loss: Checks the packet losses experienced by the applications under a given rate TCP throughput: Checks the throughput improvements for applications when the protocol decides the rate Performance of application: Checks the performance in terms of time for an FTP download
The Table 3 shows an overall summary of performance. It is clearly visible that transport protocols perform better with PMIPv6 than MIPv6. UDP on HSDPA and WLAN shows an increased available throughput of around 11% while TCP shows differing performance gains (24% for HSDPA and 51% for WLAN). This difference of performance gains is due to the TCP mechanisms being activated when WLAN starts loosing
packets. In TCP, the channel is loaded and this loading results in a sever packet loss in WLAN. Table 3 Performance summary Statistic
HSDPA
WLAN
MIPv6
PMIPv6
MIPv6
PMIPv6
UDP Throughput
837Kbps
933Kbps
2590Kbps
2890Kbps
UDP Packet Loss
15%
5.1%
13%
2.9%
TCP Throughput
1000Kbps
1240Kbps
2800Kbps
4240Kbps
FTP Download (28MB file)
185 sec
163 sec
57 sec
51 sec
Figure 3. and Figure 5. shows the performance of a 15 second window of UDP over HSDPA and WLAN, respectively. The tests show that the performance of UDP is the most stablest in HSDPA. But in both cases, PMIPv6 performs better than MIPv6.
providers. PMIPv6 possesses nice features over MIPv6 and other network based mobility management protocols. The advantages of PMIPv6 include the freeing up of the mobile host in doing any mobility related activities, the reduced signaling traffic volume and no tunneled packets in the access network. The new standard of PMIPv6 as per RFC 5213 has been implemented with basic functionality to test and evaluate the performance. The performance of the PMIPv6 implementation has been measured and compared with MIPv6. The proposed improvements in PMIPv6 regarding the increase of available bandwidth over the air interface have been measured. The measurements have been made over WLAN and HSDPA connections. The performance results have shown improvements over MIPv6 displaying its suitability for network based mobility management scenarios.
Figure 5. UDP Performance over WLAN Figure 3. UDP Performance over HSDPA
Figure 6. TCP Performance over WLAN Figure 4. TCP Performance over HSDPA
The TCP performance of the 2 protocols in HSDPA and WLAN is shown in Figure 4. and Figure 6. , respectively (15 second window). TCP mechanisms affect the throughput in both WLAN as well as HSDPA. But, when considering both protocols, PMIPv6 still performs better than MIPv6. VI.
CONCLUSION
The work evaluates the benefits of using PMIPv6 which is being considered as a possible candidate protocol for future mobility management for 3GPP and non-3GPP service
REFERENCES [1] [2] [3]
[4] [5] [6]
D. Johnson, C. Perkins, Mobility Support in IPv6, RFC 3775, June 2004 S. Gundavelli, K. Leung, V. Deverapalli and B. Patil, Proxy Mobile IPv6, RFC 5213, IETF, August 2008 Katsutoshi Nishida, et al, Implementation and Evaluation of a Network Controlled Mobility Management Protocol (IP2MM), IEEE Communication Society, 2005 R. Koodli, Ed. Fast Handove for Mobile IPv6, RFC 4068, July 2005 Julien Abeille, et al, MobiSplit: a scalable approach to emerging mobility networks, December 2006 V. Devarapalli, R. Wakikawa, Network Mobility (NEMO) Basic Support Protocol, RFC 3963, January 2005
