Transcript OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE

OPTIMIZING MOBILITY
MANAGEMENT IN FUTURE IPv6
MOBILE NETWORKS
Sandro Grech
Nokia Networks (Networks Systems Research)
Supervisor: Prof. Raimo Kantola
1
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Agenda
• The Internet Protocol (IP)
• The Internet Protocol version 6 (IPv6)
• Mobile IPv6
Background [4sl]
(I will go through these
few slides quickly!)
• Enhancements to Mobile IPv6
• Localized Mobility Management (LMM)
Foundation [5sl]
• Hierarchical LMM models
• Problem Statement
• A Non-hierarchical (flat) LMM model
• Evaluation
• Main Results
Problem Identified! [1sl]
Core of the thesis
(own contribution)
[10sl]
• Own Contribution
• Summary, Main Conclusions, Open Issues, and Further Work.
2
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Background - The Internet Protocol (IP)
• IP is a connectionless network layer protocol (each datagram is
routed independently on a hop-by-hop basis).
• IP is used to interconnect virtually any link-layer technology.
• IP is designed over an end-to-end paradigm: intelligence is in the
end systems, not in the network! The main intelligence in the
network are the routing protocols which maintain routing tables
up-to-date (used to determine the next hop towards a datagram's
destination).
• IP is stateless -> end-to-end connections do not rely on any state
being kept in some entity inside the network. As soon as some
per-connection states are introduced in some network entity, then
the connection will share the fate of that network entity.
3
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Background - The Internet Protocol version 6 (IPv6)
• Main driver: exhaustion of IPv4 addresses (different estimates regarding
when this will happen)  IPv6 extends the address space from 32 bits to
128 bits (we all heard this story...).
• IPv6 uses a streamlined header format  base header + extension
headers for optional fields  faster processing inside routers.
• IPv6 optimizes hierarchical addressing  address hierarchy is based on
the common occurrence of multiple levels of ISPs  smaller routing
tables on backbone routers (currently > 100k entries).
• IPv6 supports stateful (DHCP) and stateless (no server required)
address autoconfiguration.
• IPv6 requires the support of standards based security (IPSec).
• IPv6 flow-lables allow the provision of QoS for different traffic flows.
• IPv6 uses a new Neighbor Discovery mechanism (which replaces ARP).
4
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Mobile IPv6 [1/2]
• Benefits of mobility at the IP layer
- demand for data traffic is increasing and much of this data is
natively IP data.
- wide range of access technologies (wireless and wired) 
Mobility at the IP layer allows for a unified mobility
management mechanism across these technologies
• PROBLEM: IP addresses have a dual functionality:
- IP addresses provide a means of addressing a host
(interface). IP addresses are hierarchical <network ID>+<host
ID>  allows route aggregation (leading to smaller routing
tables)  if a Mobile Node (MN) becomes attached to a new
router, then it must change its IP address
- IP addresses are used as a means of identification at the
transport layer (e.g. TCP uses IP address + port number to
identify a session)  IP addresses should not change
5
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Mobile IPv6 [2/2]
• Mobile IPv6 provides a solution to the identified problem by:
- assigning two IP addresses to a mobile node:
+ a static Home Address (HoA)  used for identifying a MN
+ a dynamic Care-of Address (CoA)  used to maintain routability
towards the MN's most recent point of attachment to the network
- defining a new network entity (Home Agent, HA) which homes the
HoA, and maintains a mapping (binding cache) between the HoA and
CoA of MNs which are outside of the Home Network
- these mappings are also maintained by Correspondent Nodes (CNs)
which have an active session with a MN and are maintained by using
Binding Updates (BUs) whenever there is a change in CoA.
- packets from a new CN are addressed to the MN's HoA  routed to
the MN's Home Network  intercepted by the HA  forwarded
(tunneled) to the MN's CoA (based on the mapping inside the HA's
binding cache). After this the CN will receive the MN's CoA and further
packets are routed directly using the MN's CoA.
6
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Enhancements to Mobile IPv6
• Target: Provide scalable, seamless ( = lossless + fast) handovers
at the IP layer, particularly for real-time traffic.  the base Mobile
IPv6 is only providing a foundation for this.
• Enhancements to the base Mobile IPv6 mechanisms:
- Fast Handovers for Mobile IPv6 (FH)  speeds IP handovers by
allowing a MN to start the process of obtaining a new CoA prior to
disconnecting from the old Access Router
- Context Transfers (CT)  Access Routers (AR) need to maintain
several per-MN contexts which require several RTTs over the air to
be set up. CT allows these contexts to be transferred between ARs
instead of being re-established after each IP layer handover.
- IP Paging  allows a MN to go dormant (saves battery
consumption) without loosing reachability
- Localized Mobility Management (LMM) mechanisms  see next
slide
7
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
LMM: Problem Scope
• Possible issues related to the mobility management
mechanism in the base Mobile IPv6:
8
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Hierarchical LMMs [1/2]
• Introduce additional level/s of hierarchy
by introducing nodes topologically closer
to the MN which handle mobility within a
specified region, locally.
- Hierarchical Mobile IPv6 proposes to
have one node (or local mobility agent,
similar to the HA) somewhere between
the MN and HA which delimits a region
inside which this mobility agent will act as
a signaling endpoint for the MN's MIPv6
signaling. Signaling towards the HA/CNs
only needs to take place when the MN
transits to a region delimited by a
different local mobility agent.
Centralized location
information (state) required for
all DL packets towards MN
9
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
CN
HA
Local Mobility
Agent (LMA)
AR
AR
MN
Centralized location
information (state) required
only for first (few) DL packets
towards MN
Hierarchical LMMs [2/2]
- Mobile IPv6 Regional
Registrations extends this by
allowing all the routers below the
root local mobility agents to be
capable of acting as mobility
signaling endpoints. In the
optimal case the signaling would
thus only need to propagate up to
the crossover router (lowest
common router between old and
new path to MN).
Centralized location
information (state) required for
all DL packets towards MN
Centralized location
information (state) required
only for first (few) DL packets
towards MN
10
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
CN
HA
Local Mobility
Agent (LMA)
AR
AR
MN
Problem Statement
• Some problems with the hierarchical models:
- downlink packets are routed using the centralized states inside
local mobility agents instead of using standard routing table
entries  fault tolerance of IP is compromised!
- the downlink traffic needs to traverse a pre-defined path through the
local mobility agents containing the location information of the MN 
this becomes particularly unnatural for two communicating MNs
residing under the same local mobility agent, in which case the traffic
does not follow its natural shortest path.
- the MN should maintain a security association with every local
mobility agent which maintains a binding for the MN.
- require the advertisement of different domains using broadcast
messages over the bandwidth limited air interface
• Proposal: Design something simpler  solve the LMM problem using a
different perspective (see next slide)
11
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Proposal - Flat LMM
• Aim: Study the problems identified in slide 8 and formulate an
alternative approach which does not introduce the drawbacks
identified in slide 10.
• Outcome: Non-hierarchical (flat) LMM
• Outline of the proposed mechanism:
- MIPv6 e-2-e signaling latencies are tackled using temporary forwarding tunnels between the MN's
old and new Access Routers  short lifetime - just enough to allow the new
location information to reach the peer nodes (typically < 1 s)
- Signaling overhead and processing overhead are tackled by introducing a constant
(MIN_MIP_BU_INTERVAL) which defines the minimum interval between two BUs sent at the
MIPv6 layer  BUs exceeding this interval are sent as usual  other
BUs are sent to an Anchor Access Router which maintains a
mapping between the MN's old and new CoAs. This allows 'erratic'
MN transitions to be handled at the network edge and only 'stable'
transitions are notified e-2-e.
12
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Flat LMM - comparison
• Flat-LMM is not based on static local mobility areas but is
instead tightly coupled with the dynamic behavior of MNs. If a
MN is handing over at a slow rate, then no action is taken to limit
the frequency of MIPv6 signaling (note that erratic handovers
are typically very local in time, e.g. turning around corners, etc...)
• No states need to be kept in centralized states. All routers
(except for the Access Routers) are standard IP routers 
robustness of IP is retained.
• Forwarding states are only maintained at the Access Routers
(fully distributed) and are only required for a very short interval
after a handover. At all other times routing is based on standard
routing tables.
• The forwarding tunnels can be based on mechanisms defined by the
Fast Handovers protocol  additional functionality is minimal.
13
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Flat LMM
HA
Centralized location information
(state) required for delivering THE
FIRST FEW PACKETS AFTER A
HANDOVER
DL path SHORTLY AFTER A
HANDOVER
AR
DL path AFTER THE NEW
LOCATION INFORMATION
HAS PROPAGATED END-TOEND
MN
14
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
CN
Evaluation
• Aim: To evaluate the feasibility of the flat-LMM proposal
• Tools: Analytical models (assisted by some basic simulations (NS-2) where applicable) +
some measurements from prototypes when available.
• Criteria:
1) quantify the network signaling overhead caused by MIPv6 signaling
2) determine the processing overhead caused by MIPv6 signaling
1+2)  what action should be taken to reduce MIPv6 signaling
frequency (at the MIPv6 layer)?
3) what are the tunneling requirements introduced by the flat-LMM
mechanism?
15
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Main Results [1/4]
m = 2x106 subscribers
Fluid flow model:
r is the rate of flow of MNs out
of a region
 is the density of MNs inside
the region of interest
V is the average velocity of
MNs inside that region
L is the length of the region's
boundary.
Result:
~22,000 BUs/s at the HA
 ~18Mbits/s UL for BUs and
~18Mbits/s DL for BACKs.
16
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Main Results [2/4]
• On an INTEL Pentium MMX 200 Mhz the processing cost of a
Mobile IPv6 BU has been measured to be ~ 0.3 ms 
processing capacity of c.a. 3k BUs/s  a 2Ghz processor should
be capable of processing the ~20k BUs/s calculated in the
previous slide.
• Conclusions:
- Processing and signaling overhead are not appear to be as
drastic as expected.
- Reducing the 'erratic' handovers (as is done by the flat-LMM)
should be enough to maintain processing and signaling
overhead at the Mobile IP layer at a sustainable level
17
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Main Results [3/4]
• Individual mobility modeled using the
'random waypoint algorithm' provided by
NS-2.
• Assumptions:
- square cells (for simplicity)
- 50 MNs were simulated (due to long
simulation time and large log files)
• Results (summarized):
18
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Main Results [4/4]
• Tunneling requirements  Once again using the fluid flow model,
and the results shown in the previous slide the following
estimates was obtained:
- the amount of simultaneous forwarding tunnels in the flat-LMM when compared to the
Fast Handovers mechanism is increased by a factor of ~3.2 if MIN_MIP_BU_INTERVAL = 5
seconds or ~1.7 for MIN_MIP_BU_INTERVAL = 3 seconds.
- the amount of simultaneous forwarding tunnels for an Access Router serving ca. 2,000
MNs will be ~250 if MIN_MIP_BU_INTERVAL = 5 seconds or ~130 for
MIN_MIP_BU_INTERVAL = 3 seconds.
19
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
My Contribution
• Main contributions:
- Developed the non-hierarchical Localized Mobility Management model including a
detailed description of potential implementation options based on different underlying
protocol functionalities.
- Evaluation of the proposed non-hierarchical Localized Mobility Management model
using analytical models (mainly).
20
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Summary, Main Conclusions, Open
issues and Further Work
• Summary/Main Conclusions:
- Mobile IPv6 may have an important role in future mobile networks.
- Mobile IPv6 does not solve everything. Enhancements are required.
- Hierarchical LMM models exhibit some drawbacks particularly w.r.t. robustness.
- Non hierarchical LMM proposed. The main concern is the introduction of "horizontal" traffic at the edge of the
network.
• Open issues:
- Study is based on draft-15 of Mobile IPv6. Draft-16 is due to be out soon. Major changes are expected in order to
overcome the security concerns associated with draft-15.
• Further Work:
- More rigorous evaluation study based on second order models and refined assumptions.
- Investigate the impact of future revisions of the Mobile IPv6 draft (soon-to-be RFC?)
21
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002
Questions?
22
SANDRO GRECH - OPTIMIZING MOBILITY MANAGEMENT IN FUTURE IPv6 MOBILE NETWORKS :: grech_120202.ppt :: 12.02.2002