MOON: A Solution of AlwaysBestConnected in Multiinterfaced Mobile Wireless Networks

1
MOON A Solution of Always-Best-Connected in
Multi-interfaced Mobile Wireless Networks

• Ling-Jyh Chen (cclljj_at_iis.sinica.edu.tw)
• Institute of Information Science
• Academia Sinica

2
Outline of the talk

• Whats the problem?
• Technologies Wireless, Mobility,
  Multi-interfaces
• End users Always-Best-Connected (ABC)
• Why is it challenging?
• Support seamless handover
• Determine the best connection
• Survive when temporarily disconnected
• Be compatible with current Internet architecture
• Whats our solution?
• Multi-interfaced Opportunistic Overlay Network
  (MOON)

3
1. Seamless vertical handoff

• Definition
• Horizontal Handoff
• Occurs when the user switches between different
  network access points of the same kind.
• e.g. Handoff among 802.11 APs.
• Vertical Handoff
• Involves two different network interfaces which
  usually represent different technologies.
• e.g. Handoff from 802.11 to 1xRTT (CDMA 2000).

4
Seamless handoff
A seamless handoff is defined as a handoff scheme
that maintains the connectivity of all
applications on the mobile device when the
handoff occurs.
Challenges maintaining application sessions
(e.g. TCP)
5
Seamless Handoff

• Two goals low latencies and few packet losses
• Related Work
• Network Layer Approaches
• MIPv4, IPv6
• Upper Layer Approaches
• End-to-End Approaches (e.g. Dynamic DNS)
• New Session Layer Protocols (e.g. MSOCKS)
• Transport Layer Protocols (e.g. TCP-MH and SCTP)
• Middleware Approach (e.g. USHA)

6
Universal Seamless Handoff Architecture

• USHA a simple and practical handoff
  solution. (MSAN05)

NAT server
NAT Server
All packets are encapsulated and transmitted
using UDP
Applications are bound to the tunnel and
transparent to the handoff.
7
Experiment Scenario

• Ethernet 802.11b

8
Experiment (1) LOW_to_HIGH
9
Experiment (2) HIGH_to_LOW
10
Remaining issues

• Automatic vs manual handoff
• Centralized vs distributed solution
• Soft vs hard handoff
• Support Internet MANET

11
2. Intelligent Handoff Manager

• Problem
• Determining when to handoff to another interface
  is a complex decision.
• The decision may be based on various factors such
  as
• Link Capacity (speed)
• Cost
• Power Consumption
• Our Solution
• Smart Decision Model (SDM) (ANWIRE04)

12
Smart Decision Model
13
Score Function

• SD deploys a Score Function to calculate a score
  for every wireless interface
• Handoff target device is the network interface
  with the highest score.
• Score Function
• wj weight of factor j
• fj,i normalized score of interface i of factor
  j
• The equation is thus equivalent to
• where e Expense, c Link Capacity, p
  Power Consumption.

14
Score Function Breakdown

• Expense
• Link Capacity
• Power Consumption
• Note
• The coefficients a , ß , ? are determined by user
  preference.
• Inverse functions are used to bound results from
  0 to 1.
• M Maximum bandwidth requirement demanded by the
  user.
• Link capacity is calculated using
  CapProbebecause advertised link speed is seldom
  achieved due to link congestion or bad link
  quality.

15
Remaining issues

• Naming (auto-configuration)
• SDM in data link layer, network layer, or
  application layer?
• Cross-layer integration

16
3. Delay/disruption tolerant networks
1 satellite
2 dialup
Receive data from Ann
Send data to Brandy
3 memory stick on motorcycle
17
DTN Background

• DTN new network architecture for scenarios with
  intermittent connectivity
• Interplanetary communications,
• Communications between remote villages (example
  on previous slide),
• Communications with mobile users, etc.
• Common DTN characteristics
• Data is aggregated into bundles at intermediate
  node where connection breaks up.
• Bundles await the re-establishment of
  connectivity
• Bundles are sent in steps from one disconnect
  point to the next until they reach destination
  (rather than using a single end-to-end file
  transfer).

18
Our DTN Solution

• ASOS Ad-hoc Storage Overlay System
• Proposed DTN approach for MANETs
• Peer-to-peer model
• Data are transmitted end-to-end directly, if a
  path exists.
• Otherwise, data are stored in the overlay in a
  redundant and distributed fashion.

19
ASOS Example
4 ASOS data delivery
D
2 Network partitioning
ASOS overlay
1 Conventional end-to-end connection
3 Undeliverable data replicated to reachable
ASOS peers
S
20
Performance Evaluation Vehicle Scenario

• ASOS implemented in QualNet
• Virtual Track (VT) mobility model
• Battlefield scenario
• 7 switch stations
• 4 tank platoons, with 5 tanks each
• 5 HUMVEEs
• 5 relay stations
• Traffic
• CBR, 5 sources to one destination (command post),
  all moving

21
Instantaneous Throughput

• Conventional scheme
• instantaneous throughput always below input rate.
• ASOS
• Instantaneous throughput may go above input.
• Reason stored data is transmitted later when
  connectivity improves.

22
Delivery Ratio

• Overall, at the end of the data transfer, ASOS
  manages to deliver twice as much data as the
  conventional scheme

23
Remaining issues

• Storage/control overhead and optimization
• Mobility prediction
• Implementation and testbed experiments

24
4. QoS Enhancements

• Goal Providing highly agile system for mobile
  applications.

25
Related Work

• Traditional Adaptation Schemes AIMD
• TCP
• RAP, TEAR, TFRC,
• These schemes can NOT perform effectively in
  vertical handoffs Gurtov et al, MC2R04.
• Our solutions (QShine05)
• Handoff Notifications Implicit Handoff
  Notification and Explicit Handoff Notification
• Enhanced Service Agility Schemes Fast Rate
  Adaptation and Early Rate Reduction

26
Implicit Handoff Notification

• Vertical handoffs usually result in drastic
  changes in the link capacity.
• By monitoring the end-to-end capacity, one can
  identify the occurrences of vertical handoffs
• Passive capacity monitoring tools TFRC Probe
  TCP Probe

27
TFRC TCP-Friendly Rate Control

• TFRC is an equation based unicast multimedia
  streaming protocol.
• TFRC mimics the TCP long-term throughput by
  utilizing the function
• The receiver is responsible for calculating the
  loss event rate p and sending the information
  back to the sender once per round-trip time.
• The sender is responsible for adjusting its
  sending rate Tactual to be close to T.

28
TFRC Probe (E2EMON04 ComCom)

• Embed CapProbe algorithm within TFRC by sending
  two packets back-to-back every n packets
• We have evaluated the accuracy and speed of TFRC
  Probe in E2EMON04.

29
TCP Probe (GlobalInternet05)

• We have evaluated TCP Probe in GI05.

30
Implicit Handoff Notification

• A vertical handoff can be identified when a
  dramatic capacity change is observed.
• An IHN is generated when
• Cnew gt aCpre or
• Cnew lt ßCpre
• where a 5 and ß0.2.

31
Enhancing QoS using IHN

• Fast Rate Adaptation (FRA) Algorithm
• Force TCP/TFRC to enter Slow Start, instead of
  staying in Congestion Avoidance, when a vertical
  handoff from LOW to HIGH is observed.
• FRAIHN is end-to-end.
• However, IHN can not improve QoS when the handoff
  is HIGH to LOW.

32
Evaluation

• Simulation Scenario (NS-2)
• Vertical handoff from 1xRTT (150kbps) to 802.11b
  (5.5Mbps).

33
Evaluation

• TCP vertical handoff from LOW to HIGH

34
Explicit Handoff Notification

• Requirement an Intelligent Handoff Manager is
  installed.
• EHN is generated prior tothe occurrence of the
  vertical handoff.
• E.g. Smart Decision Model

35
Enhancing QoS using EHN

• When LOW-to-HIGH
• Fast Rate Adaptation (FRA) AlgorithmForce
  TCP/TFRC to enter Slow Start when a vertical
  handoff occurs.

36
Enhancing QoS using EHN

• When HIGH-to-LOW
• Early Rate Reduction (ERR) Algorithmreduce the
  sending rate in advance to prevent bulk packet
  losses.

37
Evaluation

• Simulation Scenario (NS-2)
• Vertical handoff from 802.11b (5.5Mbps) to 1xRTT
  (150kbps).

38
Evaluation

• TFRC vertical handoff from HIGH to LOW
• EHN (a) EHN when the handoff occurrs
• EHN (b) EHN one OWD before the handoff

39
Evaluation

• TFRC vertical handoff from HIGH to LOW

40
Remaining issues

• ERR for TCP
• Internet experiments
• Link capacity vs Available bandwidth?
• Monitor other network properties (e.g. buffer
  size, delay, topology, )

41
Ongoing work

• We propose a new solution called Multi-interfaced
  Opportunistic Overlay Networks (MOON).

42
MOON features

• Compatible with current network architecture
• Simultaneous data transmission on multiple
  network interfaces (based on smallest cost
  routing)
• Support distributed seamless vertical handoff
• Intelligent handoff manager
• Support DTN (e.g. ASOS)
• Enhancing QoS via passive network monitoring

43
Summary

• The requirements for the emerging
  multi-interfaced mobile wireless networks
• Compatible with current network architecture
• Seamless vertical handoff
• Intelligent handoff manager
• Delay/disruption tolerant networks
• QoS Enhancements
• We propose MOON to accommodate these
  requirements.
• The goal providing ABC in multi-interfaced
  mobile wireless networks.

44

• Thank You!