Seamless Handoffs in IP-Based Mobile Communication Networks

1. Seamless Handoffs in IP-Based Mobile Communication Networks NEC Europe MobiCom Group H. Hartenstein, F. Griffoul, K. Jonas, W. Pokorski, S. Schaller, R. Schmitz IPCN Paris, May 2000

2. Outline • Where and to what degree do we need IP-based micro-mobility support? • Wireless IP access scenarios • IP-based micro-mobility options • Our view (for discussion) • Seamless inter-domain handoffs with simultaneous bindings • Assumptions, ‘architecture’, results, open issues • Network-assisted handoffs

3. link link MIP MIP Overview scenarios: GSM-like networks “Introduce IP to mobile communication networks” simple hierarchy access core ME BTS BSC MSC Node B RNC 3GPP 3GPP2 R-P interface MWIF “IP in the RAN”

4. Overview scenarios: existing IP networks “add wireless access to IP networks” (office/corporate environment, campus, ISP) • IP ‘plug-and play’ base stations • no dedicated access network • have to take into account existing subnet structure • extreme case: ‘neighborhood networks’ micro movement but macro mobility! ISP1 ISP2

5. RNC HA AAA UTRAN 802.11 Coexistence/convergence of both scenarios ? • future trend: ‘routable’ RAN? (comes later) • connection between private and public networks?

6. Overview: IP-based micro-mobility methods • Re-addressing-based methods: “keep routes, change address”  use of care-of address & tunnels proposals: regional registrations, region-aware foreign agents, Dynamics, hierarchical MIPv6 etc. • Routing-based methods: “keep address, change routes”  no need for changing care-of address, no tunnels (within domain)  but “all or nothing” dilemma proposals: CIP, HAWAII

7. Which method for which scenario ? For GSM/UTRAN type of networks: • standard assumption ‘switching MA is far away’ not valid here • one additional ‘FA’ level might be needed • intra-RNC handoffs do not require new COA assignments Simple Linux test example: MA RTT (MN-MA): <5ms MA table update: 45ms (not optimized)

8. Which method for which scenario ? For addition of wireless access to existing corporate/campus IP networks: • Within a subnet there is no need for IP-based mobility management. • Re-addressing-based methods introduce ‘functionality’ only at certain locations, while routing-based approaches suffer from the ‘all or nothing’ dilemma. • Since RTTs are small in these environments, only small number of levels in the hierarchical setup are needed.

9. Which method for which scenario ? For future trends: • Routable access networks: • ‘IP up to the BS’ • MA still only some hops away, but COA assignments (or other negotiations) become bottleneck • here, routing-based methods might be the choice when the access network is a ‘dedicated’ network R-Point (e.g. Iu) IP NETWORK

10. Inter-domain handoffs • Motivation: micro-mobility solutions only work within an ‘administrative’ domain • We like to have seamless handoffs between different domains • Simultaneous bindings: unicast - multicast - unicast handoff • Finally, can we use this also within a domain? HA Domain1 Domain2

11. Smooth handoff as a resource problem Trade-off: degree of smoothness vs ‘costs’: • How many independent receiver/sender at mobile terminal? • e.g. UMTS/GPRS  WaveLAN handoff: seamless handoff is easy since one has two independent receiver/sender; only critical point is processing at MA. • e.g. WaveLAN  WaveLAN adhoc + same freq.: zero loss, zero delay! • How much bandwidth do we like to invest? • How many IP addresses? • Cell overlaps...?

12. Our assumptions: • one receiver/sender • e.g. intra-technology: • UMTS operator  UMTS operator • RNC  RNC • WLAN  WLAN (infrastructure, diff. freq.) • one ‘software radio’ • mobile terminal receives ‘control channel’ or BS ID on link level • two options: MN scans like in 802.11 or gets information from network (like in GSM)

13. MA 3 1 2 Domain1 Domain2 4 Seamless handoffs via simultaneous bindings Set up (via old BS) sending of duplicate packets to new BS. 1. send reg. req. with S=1 and new COA 2. start sending duplicate packets 3. reg. reply 4. handoff

14. More detailed: • we assume co-COA (but works also for FAs) • get BS ID from beacon/control channel • translate BS ID to IP subnet address • get new co-COA (via ‘address assignment server’, DHCP?) • send reg. request with S=1 and new co-COA • get reply  handoff • local reconfiguration: change co-COA, tunnels • ‘attach’ to new BS (e.g. unicast arp)

15. Results, implementational issues, open issues Our testbed: IPv4, Linux, Dynamics. Packet duplication (at MA with sim. bindings): in user space using DIVERT sockets (easy but with performance penalty) or modifying kernel modules Physical interruption: time needed for local reconfiguration. In our tests: 10 ms (but this is a system issue!) Open issues: • address assignment, address leasing, when do we give the address back? • multicast? • how to deal with ‘network unreachable’ at new BS?

16. MA Domain1 Domain2 Physical interruption vs packet loss ‘independent streams’ Physical interruption: 10 ms But what about packet loss? Depends on the delays of the different streams between MA and MN! relativistic effects! 

17. Network-assisted handoffs • what do we mean by ‘network-assisted’? • network decides/forces handoffs • network helps to set up ‘resources’ before actual handoff: ‘make before break’ • simultaneous bindings are network-assisted • why network-assisted handoffs? • traffic considerations, load balancing, interference minimization etc. • cost considerations... • seamless handoffs (vs fast handoffs)

18. Network-assisted handoffs: previous work Calhoun/Kempf proposal: FA assisted handoff, defines a handoff request message. HA 1. binding udpate 2. handoff request 1 FA FA 2

19. Network-assisted handoffs: our view • de-couple handoff support from FA functionality • send ‘request’ over old BS and use sim. bindings • otherwise one has to process two streams independently at the MN • in the case of a breakdown of the connection to the old BS, the MN is the first to notice and has to find a new BS + has to send registration request • handoff request only necessary when network wants the handoff or MN not aware of its options • network layer - link layer interworking? • does the network need ‘measurement’ reports? • if yes, how are they transported?

20. Summary • presented our view of the current micro-mobility landscape • presented experiments for inter-domain handoffs using simultaneous bindings • set up everything via old BS • physical interruption is small (10 ms) • simultaneous bindings should be kept as an option • presented ideas on a more general ‘network-assisted handoff’ framework

21. References • 3GPP Technical Report 23.923 v1.0.0, Combined GSM and MobileIP Mobility Handling in UMTS IP CN, October 1999 • Y. Xu (ed.), Mobile IP Based Micro Mobility Management Protocol in The Third Generation Wireless Network, internet draft, work in progress, March 2000 • MWIF see http://www.mwif.org • E. Gustafsson, A. Jonsson, C. Perkins, Mobile IP Regional Registration, internet-draft, work in progress, March 2000 • S. F. Foo, K. C. Chu, Regional Aware Foreign Agent (RAFA) for Fast Local Handoffs, internet draft, expired, November 1998 • Dynamics HUT Mobile IP see http://www.cs.hut.fi/Research/Dynamics • C. Castellucia, A Hierarchical Mobile IPv6 Proposal, Raport technique no 0226, INRIA, November 1998 • A. Campbell et al., Cellular IP, internet draft, work in progress, Oct. 1999 • R. Ramjee et al., IP micro-mobility support using HAWAII, internet draft, work in progress, June 1999 • another proposal employing sim. bind.: K. El Malki, H. Soliman, Hierarchical Mobile IPv4/v6 and Fast Handoffs, internet draft, work in progress, March 2000 • J. Kempf, P. Calhoun, Foreign Agent Assited Hand-off, internet draft, work in progress, January 2000 • a general paper on handoffs: N. Tripathi, J. Reed, H. VanLandingham, Handoff in Cellular Systems, IEEE Personal Communications, December 1998, pp. 26-37