Fast Intra-Network and Cross-Layer Handover (FINCH) for WiMAX and Mobile Internet

1

Fast Intra-Network andCross-Layer Handover (FINCH)

for WiMAX and Mobile InternetJ.-H. Yeh, J.-C. Chen, and P. Agrawal

IEEE Transactions on Mobile Computing, vol. 8, no. 4, pp. 558-574, 2009.

2

Outline • Introduction• Related Work• Design Principles• Proposed FINCH• Performance Analysis• Numerical Results• Conclusion

3

Introduction • 802.16 standardizes physical (PHY) layer and Media Access Control

(MAC) layer only.• To build a complete system, higher layers are necessary.

– One of the major objectives of WiMAX Forum is to promote conformance and interoperability of the IEEE 802.16 standards.

• The Access Service Network (ASN) – provides radio access to WiMAX subscribers.– consists of one or more ASN Gateways (ASN GWs) and Base Stations (BSs).

• ASNs are connected by Connectivity Service Network (CSN)– provides Internet Protocol (IP) connectivity services.

4

• Mobile IP (MIP, IETF RFC 3344) is adopted by WiMAX Forum. – The Home Agent (HA) of a Mobile Station (MS) is located in the CSN of the

MS’s Home Network Service Provider (H-NSP). – ASN GW supports the Foreign Agent (FA) functionality.

• intra-ASN mobility no need to update MS’s care-of-address (CoA). • inter-ASN mobility the MS needs to update its CoA.

5

• MIP is simple and effective – to deal with interdomain mobility management in the network layer.

• MIP’s deficiencies– frequent location update, – long handover (HO) delay, and – long end-to-end latency.

• route optimization has been proposed, – can result in latency, which is highly variable.

– may not be suitable for intradomain mobility • especially for real-time services

• IPv6 might be more efficient than IPv4, – it is not widely deployed.

• this paper only focuses on the IPv4 over Ethernet-like link model [6]– [6] J. Jee, S. Madanapalli, and J. Mandin, IP over IEEE 802.16 Problem Statement and Goals, IETF RFC 5154, Apr. 2008.

– the Address Resolution Protocol (ARP) can incur significant delay for both packet delivery and HO

6

The proposal• for interdomain mobility (inter-CSN mobility)

– use MIP only. • for intradomain mobility (intra-CSN mobility)

– a new protocol, Fast Intra-Network and Cross-layer Handover (FINCH),• which can achieve fast HO, especially for real-time services.• paging extension is designed

– to conserve the energy of MS and – reduce the signaling overhead for location update

7

Outline • Introduction• Related Work• Design Principles• Proposed FINCH• Performance Analysis• Numerical Results• Conclusion

8

Related Work• Interdomain

– MobileIP, SessionInitiateProtocol, HostIdentityProtocol

• Intradomain– Tunnel-based

• HierarchicalMIP– E. Gustafsson, A. Jonsson, and C. Perkins, “Mobile IPv4 Regional Registration,” IETF Internet draft, work in

progress, Oct. 2006. (RFC 2007)

– Host-specific-routing-based• Cellular IP

– A. T. Campbell, J. Gomez, S. Kim et al., “Design, implementation, and evaluation of cellular IP,” Personal Communications, IEEE, vol. 7, no. 4, pp. 42-49, 2000.

• Handoff-Aware Wireless Access Internet Infrastructure (HAWAII)– R. Ramjee, K. Varadhan, L. Salgarelli et al., “HAWAII: a domain-based approach for supporting mobility in wide-area

wireless networks,” TON: IEEE/ACM Transactions on Networking, vol. 10, no. 3, pp. 396-410, 2002.

• FINCH– Fast HOs for MIPv4

• R. Koodli, and C. Perkins, "Mobile IPv4 Fast Handovers," IETF, RFC 4988, 2007.

9

HMIP• Tunnel-based

Gateway FA

Regional Foreign Agent

10

Cellular IP• intends to minimize the usage of explicit signaling messages• a layer three routing protocol

– identifies MS with its home address and • MHs use the IP address of the gateway as their MIP care-of address.

– directly routes packets without tunneling or address conversion– Uplink packets

• sent to the gateway in a hop-by-hop manner. • each node on the path will cache

the source direction of the MS.– Downlink packets

• be forwarded back to the MS with the routing caches.

– When there is no data to send, MS must send a special IP packet toward the gateway to indicate its current location.

11

HAWAII• Use MIP to handle macro-mobility (interdomain)• Network nodes maintain mobile-specific routing entries on the

legacy routing tables.• MS creates, updates, and modifies the location information with

explicit signaling messages.

• The forwarding scheme – buffers packets forwarded to the

old access point and redirects them to the new access point.

• the nonforwarding scheme – drops the packets sent to the

old access point. (?)

12

Design Principles• Propsed Fast Intra-Network and Cross-layer Handover

(FINCH)– Fast HO– Cross-layer– Scalability

• avoid using centralized nodes– Paging support– Timely for deployment– Flexibility

• not be limited to tree-based topology only

13

Outline • Introduction• Related Work• Design Principles• Proposed FINCH• Performance Analysis• Numerical Results• Conclusion

14

Mobility Management in WiMAX• when an MS leaves its home network, the HA tunnels packets to the

MS’s current anchor ASN GW, which is essentially the FA

– further tunnels the packets to the MS• If the serving BS and target BS belong to different IP subnets, the MS

needs to perform L3 HO – acquire an NCoA and register the NCoA with the HA

• problems– suboptimal problem may happen in a nontree network topology– most of the protocols operate on or above the IP layer.

• They do not address link-layer mobility.– The broadcast-and-reply nature of ARP wastes bandwidth and causes extra

latency– ARP messages may wake up the MSs in sleep/idle mode

15

Cross-layer Design• a special table-lookup technique for both link layer and IP

layer to update the location. – location updates in the link layer and IP layer are coupled together.

• Consequently, ARP is no longer necessary

• The mobility management and packet routing within the domain are done by replacing the necessary routing table and bridging table with a Forwarding Table (FT).

• assume that all nodes have L3 functionality– they are capable of processing IP packets.

16

17

• When MS 1 enters the CSN-2 domain

18

Handover and Location Update

19

Characteristics • All source nodes (in the domain) know the new location of the MS

after the location update is complete. • The triangular routing in MIP is eliminated (Lee, “partially”)

– no need for route optimization.

• the new path may be the one with least congestion or shortest path– the new forwarding address is updated by the first arrived frame

• Not necessary to trigger L3 HO after L2 HO is done if it is an across IP subnet HO.– The proposed scheme deals with HOs and location update and leaves

address configuration to other protocols.

20

• Breaks the layer structure– eliminates the redundancy of doing two HOs and location

updates in both link and IP layers.

• the FT needs to maintain a list of all terminals in the domain. – Each domain does not expect to support millions or billions of

terminals.– The table searching can also be implemented in hardware

21

Paging Extension, P-FINCH

• The proposed P-FINCH is also compatible with the mobile WiMAX standards.• P-FINCH can significantly reduce signaling overhead and efficiently conserve the energy of

MSs.

• Scalable and robust– paging is initiated by each PC instead of a centralized paging initiator– Comparing with the protocols that have only one single node to buffer packets

PG: paging groupPC: paging controller

22

Outline • Introduction• Related Work• Design Principles• Proposed FINCH• Performance Analysis• Numerical Results• Conclusion

23

Performance Analysis• Handover Latency and Packet Loss• Location Update Cost• Over All Cost• Energy Conservation

24

Handover Latency and Packet Loss• Handover latency– DL2 is the L2 link switching delay

– DIP is the IP connectivity latency. • the duration of IP layer movement detection, • IP address acquisition, and configuration

– DLU is the location update latency. • the latency for binding update and • the latency to forward packets to the new IP address.

25

200 ms

200 ms

200 ms

200 ms

200 ms

10 ms

10 ms

15 ms

Handover latency Packet loss

26

Location Update Cost

• ρ the MS call-to-mobility ratio (CMR). defined as λ/μ,• U the average cost of location update to its HA in MIP.• S the cost for setting up a single link when the intradomain mobility management protocol sets up the

path in the intradomain.• A the cost of ARP operations.• L the cost for setting up the direct connection between the NAR and PAR in F-MIP.

27

U=10000L=500S=500A=1000V=200

28

Overall Cost• the overall cost adds up location update cost and packet

delivery cost.

• M the packet delivery cost of MIP.• F the packet forwarding/routing cost in the intradomain (or CSN).• T the additional re-encapsulation and decapsulation cost of MIP, F-MIP, and

HMIP.• B the cost for buffering packets at NAR in F-MIP.

29

30

Energy Conservation

31

Outline • Introduction• Related Work• Design Principles• Proposed FINCH• Performance Analysis• Numerical Results• Conclusion

32

Handover latency and packet loss

33

Location Cost and Overall Cost

• when P-FINCH is applied, the cost is reduced significantly• When the mobility rate is low, the cost of HMIP is larger than that of HAWAII and

MIP due to the hierarchical tunneling cost• the paging extension should be turned on when CMR is less than 1

mobility lowhigh

17%

34

Energy Conservation

mobility lowhigh

35

Conclusion • MIP is adopted as the mobility management protocol by WiMAX Forum.

– cannot support HOs well when mobile nodes move frequently and/or when the coverage area of a subnet is small. – even exaggerated for real-time services,

• which require very fast HOs in mobile WiMAX networks.

• a fast HO protocol, FINCH, for intradomain (intra-CSN) mobility management is proposed– discussed with examples– The analytical models and extensive simulations show that it can support fast and efficient link layer and

intradomain HOs.• reduces location update cost and overall cost because of the cross-layer design

• a scalable paging extension for FINCH, P-FINCH, is proposed– analysis also shows that it significantly reduce the signaling overhead and energy consumption

• if the size of the paging area is well configured.

• Comparing with MIP, the proposed FINCH does NOT – need IP encapsulation and – have triangular routing problem.

• also reduces the overhead caused by registering CoA with the HA• the overhead and latency in interfacing conventional mobility management protocols in the two layers

are eliminated– By unifying the mobility management in layer 2 and layer 3,

36

comments• Does ARP really has to propagate to the entire intra-CSN?

• The analysis is complete.