Feb. 21, 2008AsiaFI School on Mobile and Wireless Networks1 Enhancing IP Mobility Handover Management toward Future Internet Koshiro Mitsuya Graduate School.

Feb. 21, 2008 AsiaFI School on Mobile and Wireless Networks 1

Enhancing IP Mobility Handover Management toward Future Internet

Koshiro MitsuyaGraduate School of Media and Governance,

Keio University

WIDE Project (InternetCAR, Nautilus proj.)

[email protected]


Outline

Introduction of new handover concepts Flow-oriented smooth handover management Result of the proving test (InternetCAR testbe

d)


Background

Deployment of ubiquitous and heterogeneous wireless access networks Wi-Fi, WiMAX, HSDPA, etc. Overlapped wireless networks

Multihomed Mobile node equipped with several network interfaces switches between or simultaneously uses these interfaces

Proposition Maximally enhance the end-system to obtain the best poss

ible connectivity in such Multihomed Mobile Internet


QuickTime˛ Ç∆TIFF (LZW) êLí£ÉvÉçÉOÉâÉÄ

Ç™Ç±ÇÃÉsÉNÉ`ÉÉÇ å©ÇÈÇΩÇflÇ…ÇÕïKóvÇ≈Ç∑ÅB

Handover by definition

The process of switching the point of attachment (either L2 or L3) to the Internet

appli.

IP

Transport

appli.

NIC NIC

AP AP AP

AR AR


New Handover Concepts Per-flow Handover

Each packet determines which path (interface) should be used to transmit the packet

Policy-enabled Handover The decision is the results of the comparison of the network conditions wi

th user demand Seamless Handover

The decision dynamically changes due to e.g. the mobility Switching without service interruption is required

Adaptive Handover The switching also causes network conditions to change drastically Some degree of adaptation (e.g. bandwidth smoothing) improves the use

r experience


Discussion

Issue 1: Roaming between cell without service interruption (seamless horizontal handover)

IP Mobility (MIPv6, NEMO) To support traditional applications Upper layer mobility requires applications to be modified Problems: delay for L2 handover(A), L3 configuration(B,C,D), and CoA regist

ration(E)




Discussion (cont.)

+ Fast Handovers (FMIPv6) There are many optimizations (below) but only FMIPv6 optimize

(B,C,D) at once L2 handover, L3 configuration, CoA registration, F-HMIPv6

Open issue: We cannot avoid packet loss due to the single CoA limitation Inter-layer instructions is necessary to perform the fast handover Link layer information is media dependent


Discussion (cont..)

Issue 2: Roaming between different network interfaces (per-flow handover and policy-enabled handover)

IP Mobility (MIPv6, NEMO) MIPv6 only allows to register a single CoA per HoA It results in the hard handover

Multihoming Extension (MCoA) Allow to register multiple CoA No other standard exists

Open Issue: No mechanism to distribute traffic among available paths No mechanism to maintain the policy


Discussion (cont..)

Issue 3: Path switching is happened often due to the mobility, and it causes network condition to change drastically

Network Condition Adaptation can be implemented at several layers Transport: dynamic renegotiation of connection parameters Session: RSTP, RTP Application: bandwidth smoothing, effective compression

Open Issue: Each optimization technique solves a specific problem so that a

combination of several optimizations are required Each technique independently acquires the link layer info. hence

it is hard to integrate them


Flow-oriented Smooth Handover Management Concept

Separation of Path Maintenance and Flow DistributionSeparation of Path Maintenance and Flow Distribution



Path MaintenancePath Maintenance• IP path establishmentfor each interface• API to transparentlyuse the IP paths

Flow DistributionFlow Distribution• Path selection • Flow distribution• Policy management


Flow-oriented Smooth Handover Management Concept (cont.)

+ Handover Services



• Defines a set of primitives to assist handover operations

• Events (LinkUp, LinkDown, etc.)• Commands (LinkConnect, etc.)

• Each protocol solves a specific problem so that a combination of protocols is efficient to maximally enhance the performance

• Inter-layer or Inter-node interactions• The standard (abstracted) primitives help the integration


MNN CNMR PAR NAR

send L2 statusupdate

receive NAR info(L2/L3, QoS results)

perform L2 handoff

Handover Complete Notification

Handover Notification (pre-notification of link characteristics)

make a handover decision- decide NAR- make L2/L3 QoS reservation

Voice+Audio

Voice only

Fast Handoff(few ms)

Adapted to the change of bandwidthdynamically & surely

Example System Flow

Keep old binding cache for a while




System Diagram

L2 Abstraction IP Mobility

MIPv6, NEMO Optimizations

FMIPv6 (with proposed binding cache mngmnt)

Multiple Care-of Address Flow Distribution Adaptive Application Inter-layer information ex

change


Flow Distribution Policy(draft-mitsuya-monami6-flow-distribution-policy)

QuickTime˛ Ç∆TIFFÅiLZWÅj êLí£ÉvÉçÉOÉâÉÄ


List of availableHoA/CoA pair

A HoA/CoA pair


Example

QuickTime˛ Ç∆TIFFÅiLZWÅj êLí£ÉvÉçÉOÉâÉÄ



L2-LinkStatus Acquisition request for the current link status.

L2-PeerList Acquisition request for the list of possible access points.

L2-PeerFound / L2-PeerLost Indication of discovery/missing of candidate access points.

L2-PeerStatus QoS related information about a peer

L2-LinkUp / L2-LinkDown Notification that a new link is up / an connected link is down.

L2-LinkStatusUpdate Notification changes in the status of the connected link.

L2-LinkConnect / L2-LinkDisconnect Request for connection/disconnection of the specific link.

L2 Primitives(draft-irtf-mobopts-l2-abstractions)

* IEEE802.21 WG refer our works


L3L3

L2L2

TimeTime

L2-LinkToBeDown

L2-LinkConnect L2-LinkUp

L2 HandoverL2 Handover

L3 Handover PreparationL3 Handover Preparation L3 HandoverL3 Handover

RtSolPrRtSolPr

Mobile NodeMobile Node

Previous ARPrevious AR

New ARNew AR

HIHI

FBackFBack

FNAFNA

L2-PeerListL2-PeerFound/Lost

L2-PeerStatus

AP/AR DiscoveryAP/AR Discovery

FBUFBUPrRtAdvPrRtAdv

HAckHAck

PacketsPackets

FMIPv6 (predictive mode)with L2 Primitives


G.711 Traffic like

0

2

4

6

8

10

12

14

16

27.200 27.220 27.240 27.260 27.280 27.300 27.320 27.340 27.360 27.380

time (s)

sequence number

send

recv

mobilityrelatedsignal

1 packet lostdue to L2 handover1 packet lostdue to L2 handover

64kbps traffic80Byte payload packet10msec

Horizontal Handover Performance(against G.711 VoIP traffic)

802.11a, WEP, same channel802.11a, WEP, same channel


Vertical Handoverswith L2 Primitives



(A) L2-LinkStatusUpdate-A, (B) L2-LinkStatusUpdate-B, (C ) L2-LinkDown,(D) L2-Connect, (E) L ２− LinkDisconnect

Proposed


IPv4 Internet

iBurstNetwor

k

iBurstNetwor

k

KDDI/AU

Network

KDDI/AU

Network

In-vehicleIn-vehicle

Experiment Testbsd

MR

HUB

HUB

SIP server

VoIP Client

Log Analyzer

MobileNetwork

iBurst UT

EVDOW01K

Kyocera R&D CenterKyocera R&D Center

Keio CampusKeio Campus

VoIP Client

IPv6 Internet

HADTCPserver

G729(20bytes per 20ms)

• traffic dump• L2 info trace

• VoIP clients are communicating via a MR

• the MR is equipped with iBurst and EV-DO

• the MR will switches from iBurst to EVDO


RSSI History and Link Event(Make-After-Break Handover)


Received VoIP Traffic(Make-After-Break Handover)

37s 160s

If there is no inteaction, the performance is just terribule

If there is no inteaction, the performance is just terribule


RSSI History and Link Event(Make-After-GoinToBreak Handover)


Received VoIP Traffic (Make-After-GoingToBreak Handover)

16.9s

L2 and L3 configuration takes looong timeL2 and L3 configuration takes looong time


RSSI History and Link Event(Make-Before-Break Handover)


Received VoIP Traffic (Make-Before-Break Handover)

350ms

•Predictive metho (e.g. pre-configuration) is efficient•Packet loss is still observed because of the phased bandwidth allocation algorithm

•Predictive metho (e.g. pre-configuration) is efficient•Packet loss is still observed because of the phased bandwidth allocation algorithm


RSSI History and Link Event(Multihomed Make-Before-Break HO)


Received VoIP Traffic (Multihomed Make-Before-Break HO)

0ms,no packet loss


Adaptive Video Conference

Cellular

IEEE802.11b

Changing the video encoding rate dynamically


Conclusion

Flow-oriented Handover Management + Media Independent Handover Service Successfully developed the new end-system arch

itecture that enables each application to roam seamlessly and adaptively between ubiquitously deployed networks

Future work QoS guarantee Power management


Thank you for your listening!



Key Contribution

Handover Service Defined the abstracted link layer events and commands for the persiste

nt IP path maintenance Defined the abstracted link layer events and commands for the flow migr

ation FMIPv6 related

Modified the specification on the binding management Extended FMIPv6 protocol to carry other information needed for handov

er optimization Provided the world’s first open source implementation of FMIPv6 Feedback to the spec, Interoperability tests, Demonstrations


Key Contribution (cont.)

Flow Distribution Defined the policy language to describe user demands for t

he flow distribution Developed the policy exchange protocol among end and a

nchor nodes of communications Network Condition Adaptation

Developed a protocol to share the network condition among end and anchor nodes of communications

Feb. 21, 2008AsiaFI School on Mobile and Wireless Networks1 Enhancing IP Mobility Handover Management toward Future Internet Koshiro Mitsuya Graduate School.

Documents

asiafi school

flow handover

wireless networks4 handover

adaptive handover

wireless networks1

policy slide

packet policyenabled

wireless networks7 discussion