Download A Scheme to Reduce Packet Loss during PMIPv6

International Conference on Computational Sciences and Its Applications ICCSA 2008
A Scheme to Reduce Packet Loss during PMIPv6 Handover
considering Authentication∗
Seonggeun Ryu, Gye-Young Kim, Byunggi Kim, and Youngsong Mun
School of Computing
Soongsil University, Korea
[email protected], {gykim11, bgkim, mun}ssu.ac.kr
Abstract
based on IP technology. Mobile IPv6 (MIPv6) [1] from
the Internet Engineering Task Force (IETF) is standardized
for mobility management in IPv6 wireless/mobile networks.
Mobile IPv6 requires client functionality in the IPv6 stack
of a mobile node. Exchange of signaling messages between
the MN and a home agent (HA) enables the creation and
maintenance of binding between the MN’s home address
and its care-of address. Mobility as specified in MIPv6 requires the IP host to send IP mobility management signaling messages to the HA, which is located in the network.
MIPv6 is a approach of host-based mobility to solve the IP
mobility challenge. However, it takes a long time to process
handover and there is much packet loss during handover,
since there are many signaling messages via wireless link
which occurs long delay during handover.
Network-based mobility is another approach to solving
the IP mobility challenge. It is possible to support mobility for IPv6 nodes without host involvement by extending
MIPv6 signaling messages and reusing the HA. This approach to supporting mobility does not require the MN to
be involved in the exchange of signaling messages between
itself and the HA. A Mobile Access Gateway (MAG) in
the network performs the signaling with the HA and does
the mobility management on behalf of the MN attached to
the network. This protocol is called as Proxy Mobile IPv6
(PMIPv6) [2] in Network-based Localized Mobility Management (NETLMM) working group of IETF. PMIPv6 can
reduce handover latency, since the proxy mobility agent on
behalf of the MN performs handover process. That is, there
are a little signaling message via wireless link.
There is much packet loss during handover in PMIPv6,
although PMIPv6 reduces handover latency. In this paper, to reduce packet loss in PMIPv6 we propose PacketLossless PMIPv6 (PL-PMIPv6) with authentication. The
similar scheme was studied to reduce packet loss and handover latency in MIPv6, such as fast handovers for MIPv6
(FMIPv6) [3]. In PL-PMIPv6, a previous MAG (pMAG)
registers to a Local Mobility Anchor (LMA) on behalf of
a new MAG (nMAG) during layer 2 handoff. Then, the
Mobile IPv6 (MIPv6) is a presentative protocol which
supports global IP mobility. MIPv6 causes a long handover
latency that a mobile node (MN) doesn’t send or receive
packets. This latency can be reduced by using Proxy Mobile IPv6 (PMIPv6). PMIPv6 is a protocol which network
supports IP mobility without participation of the MN, and is
studied in Network-based Localized Mobility Management
(NETLMM) working group of IETF. There is much packet
loss during handover in PMIPv6, although PMIPv6 reduces
handover latency. In this paper, to reduce packet loss in
PMIPv6 we propose Packet Lossless PMIPv6 (PL-PMIPv6)
with authentication. In PL-PMIPv6 a previous mobile access gateway (pMAG) registers to a Local Mobility Anchor
(LMA) on behalf of a new MAG (nMAG) during layer 2
handoff. Then, the nMAG buffers packets during handover
after registration. Therefore, PL-PMIPv6 can reduce packet
loss than them in MIPv6 and PMIPv6. Also, we use Authentication, Authorization and Accounting (AAA) infrastructure to authenticate the MN and to receive MN’s profiles securely. We shows performance of PL-PMIPv6 through comparison of packet loss during handover of MIPv6, PMIPv6
and PL-PMIPv6.
1. Introduction
In wireless/mobile networks, mobile nodes (MN) can
change their attachment points while they communicate
with correspondent nodes (CN). Hence, mobility management is essential for tracking the MNs current locations so
that their data can be delivered correctly. IP-based mobility management is critical, since the next-generation wireless/mobile networks are anticipated to be unified networks
∗ This research was supported by Ministry of Knowledge Economy,
Korea, under the Information Technology Research Center support program supervised by the Institute of Information Technology Advancement.
(grant number IITA-2008-C1090-0801-0027)
978-0-7695-3243-1/08 $25.00 © 2008 IEEE
DOI 10.1109/ICCSA.2008.23
47
MN
p-MAG
n-MAG
LMA
AAA
MN
Layer2 Handoff
MN Detached
MN Detached Event
p-MAG
n-MAG
LMA
AAA
MN Detached
MN Detached Event
Layer2 Handoff
DeReg PBU
DeReg PBA
MN Attached
Buffering
is begun
Tunneling
DeReg PBA
PBA
MN Attached Event
Access Initiation
AAA Request
Access Accept
AAA Response
Tunneling
is begun
DeReg PBU
+ PBU at n-MAG
MN Attached
Access Initiation
MN Attached Event
AAA Request
Access Accept
PBU
Tunneling
AAA Response
Router Solicitation
PBA
Router Advertisement
Router Solicitation
Router Advertisement
Figure 2. Signaling flow of the proposed
scheme (PL-PMIPv6).
Figure 1. Signaling flow of PMIPv6 with Authentication.
the MN and acquiring its identity, will determine if the MN
is authorized for the network-based mobility management
service. For updating the LMA about the current location
of the MN, the MAG sends a Proxy Binding Update (PBU)
message to the MN’s LMA. Upon accepting this PBU message, the LMA sends a Proxy Binding Acknowledgement
(PBA) message including the MN’s home network prefix.
It also creates the Binding Cache entry and establishes a bidirectional tunnel to the MAG. The MAG on receiving the
PBA message sets up a bi-directional tunnel to the LMA and
sets up the data path for the MN’s traffic. At this point the
MAG will have all the required information for emulating
the MN’s home link. It sends Router Advertisement (RA)
messages to the MN on the access link advertising the MN’s
home network prefix as the hosted on-link-prefix. Figure 1
shows the signaling call flow for the MN’s handover from
previously attached MAG (pMAG) to the newly attached
MAG (nMAG). After obtaining the initial address configuration in the PMIPv6 domain, if the MN changes its point
of attachment, the MAG on the previous link will detect the
MN’s detachment from the link and will signal the LMA
and will remove the binding and routing state for that MN.
However, the LMA upon accepting the request will wait for
certain amount of time before it deletes the binding, for allowing a smooth handover. The MAG on the new access
link upon detecting the MN on its access link will signal the
LMA for updating the binding state. Once that signaling is
complete, the MN will continue to receive the RAs containing its home network prefix, making it believe it is still on
the same link and it will use the same address configuration
on the new access link.
nMAG buffers packets during handover after registration.
Therefore, PL-PMIPv6 can reduce packet loss than them
in MIPv6 and PMIPv6. Also, we use Authentication, Authorization and Accounting (AAA) infrastructure to authenticate the MN and to receive MN’s profiles securely. We
shows performance of PL-PMIPv6 through comparison of
packet loss during handover of MIPv6, PMIPv6 and PLPMIPv6.
The rest of the paper is organized as follows. Section 2
specifies PMIPv6 protocol as related works. Packet-lossless
PMIPv6 (PL-PMIPv6) is proposed in Section 3. Section
4 presents analytical model and the numerical results. In
Section 5, we conclude this paper.
2. Related Works
PMIPv6 is intended for providing network-based IP mobility management support to a mobile node, without requiring the participation of the MN in any IP mobility related
signaling. The mobility entities in the network will track
the MN’s movements and will initiate the mobility signaling and setup the required routing state.
The core functional entities in the NETLMM infrastructure are the Local Mobility Anchor (LMA) and the Mobile
Access Gateway (MAG). The LMA is responsible for maintaining the MN’s reachability state and is the topological anchor point for the MN’s home network prefix. The MAG is
the entity that performs the mobility management on behalf
of an MN and it resides on the access link where the MN is
anchored. The MAG is responsible for detecting the MN’s
movements to and from the access link and for initiating
binding registrations to the MN’s LMA.
Once a MN enters PMIPv6 domain and attaches to an
access link, the MAG on that access link, after identifying
3. Proposed Scheme (PL-PMIPv6)
In PMIPv6, there are schemes to reduce handover latency and packet loss [4] [5]. Fast handovers for PMIPv6 is
48
AAA
PBU
uest
AA A Req
n
Solic ita tio
tion
Access In
itia
MN
Attached
ent
Advertisem
pt
MN
Rout er
pMAG
MN
A cce
Access
MN
Detached
Even
n-MAG/AR
PBA
p-MAG/AR
DeRe g PBA
nMAG
se
AAA Respon
LMA
Home Network
Internet
Backbone
MN
Detached
Even
Router
AAA
DeReg PBU
LMA/HA
MIPv6
Handover
End
Epoch
MN
Figure 3. System Model.
Time
tlink − switching
t AAA− Auth t P − Re gistration
t RS / RA
Total Handover Latency
a scheme that only LMA exchange signaling with MAGs to
set up the fast handover [4]. However, this scheme does not
follow the order of signaling flow in PMIPv6. The scheme
in [5] is the Extended PMA (EPMA) and LPMA functionalities to reduce signaling cost for intra domain handover and
to optimize packet delivery [5]. However, this scheme does
not prevent packet loss during handover.
In this paper, we propose Packet-Lossless PMIPv6 (PLPMIPv6) with authentication, to reduce packet loss in
PMIPv6. PL-PMIPv6 follows the order of signaling flow
in PMIPv6 and reduces packet loss. Figure 2 shows signaling flow of PL-PMIPv6 during handover. After the pMAG
is aware of the MN’s detachment, it sends the DeReg PBU
message to the LMA in PMIPv6. In PL-PMIPv6 when
pMAG sends the DeReg PBU message, the PBU message
of nMAG is included in DeReg PBU message. That is, the
pMAG registers on behalf of the nMAG in advance to reduce handover latency. As a result, the tunnel between the
LMA and the nMAG is established in advance. Also, when
the nMAG receives the PBA message, it begins to buffer
packets to the MN. After layer 2 handoff, the MN sends
the RS message and receives the RA message including the
MN’s home network prefix.
In PMIPv6, we use AAA infrastructure to authenticate
the MN like in [5]. Then, the nMAG can receive the MN’s
profile securely using AAA infrastructure.
Figure 4. Handover latency of PMIPv6.
movement. When the movement and packet generation processes are independent and stationary, the PMR is given by
ρ = λ/µ. We assume that a cost for transmitting a packet is
dependent on the distance between the sender and receiver.
We define that ld is the average length of a data packet and
lc is the average length of a control packet [7].
4.1. Handover Latency
In this section, we analyze handover latency of MIPv6,
PMIPv6 and PL-PMIPv6. To analyze handover latency of
three schemes, we define that tM N,M AG , tM AG,HN and
tM AG,M AG are transfer delays between an MN and an
MAG, an MAG and a LMA, and adjacent two MAGs, respectively.
Handover latency consists of three latencies such as a
link switching latency, an IP connectivity latency and a location update latency. The link switching latency is due to
a layer 2 handoff. The IP connectivity latency is due to
movement detection and new IP address configuration after
the layer 2 handoff. An MN can send or receive packets
in nMAG after the IP connectivity. We define handover
latency of MIPv6 like as,
M IP v6
TLatency
4. Performance Evaluations
= tlink−switching + tAAA−Auth +
+ tAddr−Autoconf + tRegistration . (1)
We make a comparison of MIPv6, PMIPv6 and PLMIPv6 handover in terms of handover latency, costs and
total cost ratio. For those comparisons, we use a system
model in Fig. 3. In the system model, we evaluate performance of three schemes when an MN moves between
MAGs. We assume that a correspondent node generates
data packets destined to the MN at a mean rate λ, and the
MN moves between MAGs at a mean rate µ. We define
packet to mobility ratio (PMR, ρ) as the mean number of
packets received by the MN from the correspondent per
tlink−switching is a delay during layer 2 handover.
tAAA−Auth is a delay during authentication of an MN
through AAA infrastructure (2 · (tM N,M AG + tM AG,HN )).
tAddr−Autoconf is a delay during process of stateless autoconfiguration (2·tM N,M AG ). tRegistration is a delay during
binding update to a HA (2 · (tM N,M AG + tM AG,HN )).
In Fig. 4, we define handover latency of PMIPv6. Handover latency of PMIPv6 is following,
P M IP v6
TLatency
49
= tlink−switching + tAAA−Auth +
1600
MN
Detached
Even
se
AAA Respon
DeRe g PBA
AA A Reques
t
1200
n
Solic ita tio
Access
Initia
MN
ent
vertisem
pt
MN
Attached
Rout er Ad
tion
Packet
Buffering
Start
pMAG
t AAA − Auth
1000
800
600
400
MIPv6
Handover
End
Epoch
Packet Buffering
tlink − switching
MIPv6
PMIPv6
PL−PMIPv6
1400
Packet
Buffering
End
A cce
Access
MN
Detached
Even
g
tunnelin
P BA
nMAG
DeReg PBU inc
luding PBU at nMA
G
LMA
Packet delivery cost
Packet
Forwarding
Start
Router
AAA
200
Time
t RS / RA
0
t buffering
0
200
400
600
PMR(rho)
800
1000
Figure 7. PMR vs packet delivery cost.
Total Handover Latency
Figure 5. Handover latency of PL-PMIPv6.
Handover latency is seriously affected by accesses of
wireless link during handover. Hence, handover latencies of
PMIPv6 and PL-PMIPv6 is lower than that of MIPv6. Also,
PL-PMIPv6 reduces more handover latency than PMIPv6,
since the registration of nMAG is performed during layer 2
handoff.
500
MIPv6
PMIPv6
PL−PMIPv6
450
Handover latency (ms)
400
350
300
4.2. Total Cost during Handover
250
In this section we analyze three scheme in terms of total cost. Total cost consists of signaling cost and packet
delivery cost during handover. Signaling cost is cost of
messages for signaling, and packet delivery cost is cost of
packet transferred from a HA.
We assume that signaling and delivery cost are influenced by transmission delay. That is, signaling cost and
delivery cost of a packet consist of an average length of signal and data packet and transmission delay, respectively.
Followings are signaling cost of three scheme.
200
150
100
10
20
30
40
Wireless link delay (ms)
50
60
Figure 6. Handover latency vs wireless link
delay.
+ tP −Registration + tRS−RA .
(2)
tP −Registration is a delay during proxy binding update to a
LMA (2·tM AG,HM ). tRS−RA is a delay during exchanging
of a router solicitation (RS) and a router advertisement (RA)
messages between the MN and the MAG (2 · tM N,M AG ).
Handover latency of PL-PMIPv6 can be defined by formula 3 through Fig. 5. In this case, delay of proxy registration is reduced, since proxy registration is performed during
layer 2 handoff.
P L−P M IP v6
TLatency
M IP v6
CSignaling
= SAAA−Auth + SAddr−Autoconf +
+ SRegistration ,
(4)
P M IP v6
CSignaling
= SP −Deregistration + SAAA−Auth +
+ SP −Registration + SRS−RA , (5)
P L−P M IP v6
CSignaling
= SP −Deregistration + SAAA−Auth +
+ SP −Registration /2 + SRS−RA .(6)
SAAA−Auth is cost for process of AAA authentication
(lc · tAAA−Auth ). SAddr−Autoconf is cost for stateless address autoconfiguration (lc · tAddr−Autoconf ). SRegistration
is cost for binding update to a HA (lc · tRegistration ).
SP −Registration is cost for proxy binding update to a LMA
(lc · tP −Registration ). SRS−RA is cost for exchanging a RS
and a RA messages (lc · tRS−RA ). We have known that
signaling costs of three scheme differ little.
= tlink−switching + tAAA−Auth +
+ tRS−RA .
(3)
We present some results based on the above analysis. Figure 6 shows handover latency of three schemes. In Fig.
6, we use that tM N,M AG = 10ms, tM AG,M AG = 2ms,
tM AG,HN = 10ms, referring to [8].
50
1
5. Conclusions
PL−PMIPv6 / PMIPv6
PL−PMIPv6 / MIPv6
0.9
In PMIPv6 all packets from CNs are lost during handover. To solve this problem, we have proposed PLPMIPv6. PL-PMIPv6 can prevent packet loss during handover, since process of proxy registration is performed
in advance and then packets from CNs are tunneled and
buffered to nMAG. Also, we used AAA infrastructure to
authenticate a MN and to receive MN’s profiles securely.
Hence, security of PMIPv6 is enhanced.
We showed performance of PL-PMIPv6 through performance evaluations. PL-PMIPv6 reduced 26% of total cost
compared with PMIPv6. This shows that PL-PMIPv6 can
enhance PMIPv6 through tunneling and buffering.
Recently, PMIPv6 is in process of last call in IETF and
will be a standard document. Then, if PL-PMIPv6 is used in
addition to PMIPv6, PMIPv6 will be a more robust protocol
in NETLMM working group of IETF.
Total cost ratio
0.8
0.7
0.6
0.5
0.4
0.3
0
200
400
600
PMR(rho)
800
1000
Figure 8. PMR vs total cost ratio.
We assume that delivery cost consists of packet transmission cost and packet lost during handover. Packet delivery
costs of three scheme are followings,
M IP v6
CDelivery
M IP v6
= λ · dHN,M N · TLatency
· η,
P M IP v6
CDelivery
P M IP v6
= λ · dHN,pM AG · TLatency
· η, (8)
P LP M IP v6
CDelivery
References
[1] D. Johnson, C. Perkins, and J. Arkko, ”Mobility Support in IPv6,” RFC 3775, IETF, June 2004.
(7)
[2] S. Gundavelli, K. Leung, V. Devarapalli, K. Chowdhury, and B. Patil, ”Proxy Mobile IPv6”, draft-ietfnetlmm-proxymip6-10.txt, IETF, February 2008.
= λ · dHN,M N · tM AG,HN · η +
+ λ · dHN,nM AG · tbuf f ering . (9)
[3] R. Koodli, Ed., ”Fast Handovers for Mobile IPv6,”
RFC 4068, IETF, July 2005.
dHN,M N is delivery cost from home network to a MN.
dHN,pM AG is delivery cost from home network to a pMAG.
dHN,nM AG is delivery cost from home network to a new
MAG through the pMAG. In MIPv6 and PMIP, all packets sent from the CN are lost during handover. However,
in PL-PMIPv6 most packets are tunneled to nMAG and are
buffered. Figure 7 shows delivery costs of three schemes.
In Fig. 7 costs of PMIPv6 and PL-PMIPv6 are much lower
than MIPv6, since handover latencies are short in PMIPv6
and PL-PMIPv6. In addition, cost of PMIPv6 is the lowest
in three schemes because packets from the CN are tunneled
and buffered in PMIPv6.
We calculate total costs of three schemes using above
formulas. Then we calculate cost ratio following formulas
to compare performance of three schemes.
P L−P M IP v6
CostRatioM
IP v6
P L−P M IP v6
CostRatioP
M IP v6
[4] P. Kim, S. Kim, and J. Jin, ”Fast Handovers for
Proxy Mobile IPv6 without Inter-MAG Signaling,”
draft-pskim-netlmm-fastpmip6-00.txt, IETF, November 2007.
[5] S. Park, N. Kang, and Y. Kim, ”Localized ProxyMIPv6 with Route Optimization in IP-Based Networks,” IEICE TRANS. COMMUN., VOL.E90.B,
NO.12, December 2007.
[6] S. Ryu and Y. Mun, ”The Tentative and Early Binding
Update for Mobile IPv6 Fast Handover,” LNCS 3794,
Springer, MSN 2005, December 2005, pp.825-835.
[7] R. Jain, T. Raleigh, C. Graff, and M. Bereschinsky,
”Mobile Internet Access and QoS Guarantees using
Mobile IP and RSVP with Location Registers,” Proc.
IEEE Intl. Conf. on Communications, IEEE, Atlanta,
June 1998, pp.1690-1695.
T otalCostP L−P M IP v6
=
(10)
T otalCostM IP v6
T otalCostP L−P M IP v6
=
(11)
T otalCostP M IP v6
[8] H. Fathi, R. Prasad, and S. Chakraborty, ”MOBILITY MANAGEMENT FOR VOIP IN 3G SYSTEMS: EVALUATION OF LOW-LATENCY HANDOFF SCHEMES,” IEEE Wireless Communications,
Volume 56, Issue 1, January 2007, pp.260-270.
Figure 8 shows total cost ratios of MIPv6 and PMIPv6
about PL-PMIPv6. Through this figures we show that PLPMIPv6 reduces total cost during handover. That is, PLPMIPv6 enhances handover performance of 66% compared
with MIPv6 and 26% compared with PMIPv6.
51