Introduction to IP Mobility Jon-Olov Vatn
KTH/IMIT/TSLab
Introduction to IP Mobility Jon-Olov Vatn
KTH/IMIT/TSLab
Handover at Layer 2 and 3
CN HA AP AP R AP AP R Internet MN MN MN Layer-2
In-LAN
Layer-3
Between LANs
New IP address needed
IP Mobility not only for single interface or link technology
Internet Protocol (IP) GSM DECT UMTS 802.11 HyperLAN/2 Inter-operability handled by IP:
Internetworking rather than interworking.
Easy introduction of new link layers.
Problem with IP handover
IP address identifies an interface.
IP is associated with a link/subnet (for aggregation purposes)
IP address (partly) identifies some higher layer sessions (in particular TCP)
Changing IP address may break higher layer connections.
Approaches to IP mobility
Provide ”portability” instead of ”mobility”
Use dynamic address assignment
Don’t bother if higher layer connections are broken.
Use two IP addresses (Mobile IP)
Home address (for session identification)
Care-of address (identifies current location)
Tunneling/Routing header
Application specific solutions
SIP Mobility (Re-Invite)
Example: Mobile IPv4
Triangular routing (longer end-to-end delay) CN HA FA MN MN MN MIPv4 Introduce:
Home Agent (HA)
Foreign Agent (FA)
Mobile Node (MN)
Correspondent Node (CN)
Care-of Address (COA)
HA intercepts and tunnels packets to FA (care-of address)
Transparent to CN
MIPv4 variants
CN FA FA MN HA Bi-directional tunnels Route optimization
Need specific support at CN
Privacy problem
Bi-directional tunnels
Longer delay
Privacy
Handle security concious routers
Co-located COA
No need for FA
Use DHCP or similar to acquire COA
Mobile IPv6
CN R R MN HA IETF draft (version 24)
About to become RFC?
Only co-located COAs
No Foreign Agents
Route optimization
All IPv6 nodes (including CNs) will(?) have MIP support
Routing headers
Binding cache
Securing binding updates
MN-HA: IPSec
MN-CN: ”Return routability”
Privacy problem?
(Fall-back on bi-directional tunnels)
Mobile IP deployment
Bad signs
Chicken and egg problem:
No MIP support in hosts => Access networks are built as big LANs
Access networks as big LANs => No need for MIP
Real-time applications are UDP based and could use application specific solutions, e.g., SIP mobility.
Good signs
IPv6 deployment will imply MIPv6 deployment
If FA is co-located with NAT, it solves NAT problems.
Handles all higher layer sessions (including TCP)
More information
IETF Working groups
Links to relevant drafts and RFCs
MIPv4
http://www.ietf.org/html.charters/mip4-charter.html
MIPv6
http://www.ietf.org/html.charters/mip6-charter.html
Other relevant WGs
MIPShop (MIPv6 Signaling and Handoff Optimization)
Nemo (Network Mobility)
AAA (Authentication, Authorization and Accounting)
Comments