Title: Creating AlwaysBestConnected Multimedia Application for the 4:th Generation Wireless Systems
1Creating Always-Best-Connected Multimedia
Application for the 4th Generation Wireless
Systems
- Johan Kristiansson
- Department of Computer Science and Electrical
Engineering - Division of Media Technology
- Luleå university of technology
- Licentiate presentation 2004-11-30
Supervisor/examinator Dr. Peter
Parnes Discussant Prof. Gerald Maguire Jr.
2Agenda
- Introduction
- Papers
- Application-layer Mobility support for Streaming
Real-time Media - Providing Seamless Mobility with
Competition-based Soft Handover Management - Market-based Bandwidth Management for Distributed
Multimedia Applications - Conclusion and future work
- Discussion
3Introduction
- Rapid development of wireless networks
- UMTS/3G networks
- WiFi networks
- 49,263 public hotspots
- Ad-hoc networks, e.g. MeshNetworks
- Internet and laptop use exploding
- 50 of all laptops have WLAN support
- New types of applications
- Wearable computing
Source ) http//www.jiwire.com/ (November 2004)
44th generation wireless system
- There is not yet a universally agreed-upon
definition of 4G - for some 3G WLAN access is already 4G
- for some bandwidth gt 20Mb/s is already 4G
- Diversity of wireless networks
- WiMax, UWB etc.
- Always-Best-Connected
- Operator-centric view
- User-centric view
5Today wireless landscape
Network type n/a Capacity n/a Price n/a
Network type 802.11b Capacity 11 Mbit/s Price
Free
Network type GSM/GPRS Capacity 40.2
kbit/s Price 20 kr/Mb
Network type 802.11b Capacity 500 kbit/s Price
Free
Network type Wired Ethernet Capacity 100
Mbit/s Price Flat-rate
Network type UMTS Capacity 300 kbit/s Price 18
kr/Mb
This is a fictitious coverage map
6Motivation
- How to creating multimedia applications that can
efficiently take advantage of available bandwidth
- How to utilize bandwidth in the possible way in
the application? - How to supply the application with bandwidth in
the best possible way? - It should just work?!
- How to hide heterogeneity?
7Research problem
- Seamless mobility (Paper 1 and 2)
- How to allow applications switching (handover)
between networks? - Why and when should a handover to another network
be initiated? - To which network?
- Adaptability (Paper 3)
- How to adapt multimedia applications to available
bandwidth? - How can bandwidth be shared between media stream?
8Related 4G problems
- Radio technologies
- Power control, smart antennas etc.
- Security aspects
- Authentication, Authorization etc.
- Business models
- Billing etc.
- Network management
- Mobile multimedia applications
9Research methodology
- Exploratory research methodology
- Marratech Pro
- Middleware
- Real-life experiments
- Short time frame
10Paper 1Application-layer Mobility support for
Streaming Real-time Media
- Johan Kristiansson, and Peter Parnes
- Department of Computer Science and Electrical
Engineering - Division of Media Technology
- Luleå university of technology
- Presented at IEEE Wireless Communication and
Networking Conference (WCNC), 2004, Georgia,
Atlanta
11Introduction
- Motivation
- Many applications cannot handle multi-homing or
handle longer periods of time-outs. - Research issues
- How to create multimedia applications that can
rebind (handover) to a new network interface at
any time? - Contribution
- A new UDP socket abstraction (Resilient Mobile
Socket)
12Related work
- Mobile IP
- Requires support from the infrastructure
- Limited NAT support
- Triangular routing
- No support for multiple interfaces
- Mobile IPv6
- Stream Control Transmission Protocol (SCTP)
- Previous implementations only supported TCP
- Limited support for mobility (ADD-IP)
- NAT and general firewall problem (ALG is needed)
- Operating system support is required
- Datagram Congestion Control Protocol (DCCP)
- Limited support for mobility
- Few implementations
13Resilient Mobile Socket
- What is a socket?
- Represents an end-point of communication link to
other applications running on the network - Used by applications to send and receive packets
14Resilient Mobile Socket
- A new socket abstraction for UDP
- Each internal socket represent a carrier/operator
- Packet translation
- Both end-point must be modified
15Resilient Mobile Socket
- A new socket abstraction for UDP
- Each internal socket represent a carrier/operator
- Packet translation
- Both end-point must be modified
16Demo time!! ?
17 18Same experiment but using the RMS-enabled
prototype
19 20Results
- Can jitter or packet delay be used to detect a
disconnection?
21Discussion
- Advantages
- No support from the infrastructure is required
- Support for multiple interfaces
- Multicast support. Switch between an internal
unicast and a multicast socket - No triangular routing
- Disadvantages
- Only UDP is supported
- All application must be use RMS, i.e. all
application must be modified - No security. Connection hijacking possible.
22Paper 2Providing Seamless Mobility with
Competition-based Soft Handover Management
- Johan Kristiansson, and Peter Parnes
- Department of Computer Science and Electrical
Engineering - Division of Media Technology
- Luleå university of technology
- Published at 7th IPIP/IEEE International
Conference on Management of Multimedia Networks
and Services (MMNS), California, San Diego, 2004
23Introduction
- Motivation
- Real-time media require seamless mobility
- Research issues
- Minimize handover delay
- Ping-pong problem
- What is the best network for real-time media?
- When should a handover be initialized?
- Contribution
- A new handover strategy for application-layer
mobility
24Related work
- Architectures for minimizing the handover delay
- Cellular IP, Hierarchical Mobile IP etc.
- Traditional ways to trigger handovers
- Dwell-time, hysteresis, threshold based algorithm
- L2 information, e.g. signal-strength
- Soft handover support
- Must be proactive
- Location-aided handover
25CSHM - Problems
- When should redundancy be enabled?
- When should it be disabled?
- How to deal with duplicate packets?
- Which connection should be used after redundancy
has been disabled?
26CSHM - Solutions
- When should redundancy be enabled?
- Proactive soft handovers based on packet delay
- Packet must be sent with a regular interval
- Both end-points are involved in decision process
- If packet delay gt threshold (F) then enable
redundancy
27CSHM - Solutions
- When should redundancy be disabled?
- A dwell timer (D) is used to disable redundancy
- How to deal with duplicate packets?
- Filtering out duplicate packets using seq. nr.
- New performance metric packet contribution
- Which connection should be used after redundancy
has been disabled? - The one with the highest packet contribution
28Evaluation
- Investigate F and D relationship between packet
delay and losses, and duplicated packets - Experimental setup
- Two WiFi networks
- Two clients sending GSM audio
- Hard to repeat experiments
- Emulation
29 30Discussion
- Advantages
- Minimal handover delay
- High handover frequency (Ping-pong problem)
- Evaluate end-to-end performance
- Minimize duplicated packets
- Disadvantages
- Not suitable for all media
- Calibration of F and D is needed
- Requires access to multiple network interfaces
- Can make congestion worse
31Paper 3Market-based Bandwidth Management for
Distributed Multimedia Applications
- Johan Kristiansson, Jeremiah Scholl, and Peter
Parnes - Department of Computer Science and Electrical
Engineering - Division of Media Technology
- Luleå university of technology
- Under review
32Introduction
- Motivation
- Applications must adapt to available bandwidth
- Research issues
- Help multimedia applications better leverage
available bandwidth - Sharing/allocating bandwidth between adaptive
media - Supplying application with bandwidth
- Contribution
- A middleware based on microeconomics for managing
bandwidth
33Related work
- QoS-aware middlewares for operating system
- Agilos
- B. Li. Agilos A Middleware Control Architecture
for Application-Aware Quality of Service
Adaptations. PhD thesis, University of Illinois,
USA, 2000. - Q-RAM
- Microeconomics
- Congestion pricing
- CPU allocations in operating systems
34Microeconomy
- Microeconomy
- Consumers and producers
- Invisible hand
- Equilibrium
- Tâtonnement process
- Spot-market
35Calculating the supply
- Constraints
- Budget of the user
- Capacity of the network(s)
- smin(budgetUser, capacityNetwork)
- Congestion control protocol
- Sender-based vs receiver-based congestion control
- TCP-Friendly Rate Control Protocol (TFRC)
36Calculating the demand
37Overview of the middleware
38Evaluation
Video 25 kB/s
B
A
Audio
Video
LAN
GPRS
39Experiment 1
Introducing a new media (audio)
Price variation
Video and audio demand variation
40Experiment 2
Supply recalculation rate
Price recalculation rate
Price recalc. rate50ms
Supply recalc. rate800ms
Equilibrium
41Experiment 3
Switching between a LAN network and GPRS network
Price variation
Video demand variation
42Discussion
- Advantages
- Plug-and-play replacement of components
- Equilibrium allocations are possible
- Disadvantages
- Using a high supply recalc. rate results less
bandwith utilization - Only selling bandwidth from one network at a time
43Final remark
- Always-Best-Connected
- More than just being best connected
- Most also consider how to maximize the users net
benefit - Cannot modify all applications
- A middleware is needed!!
44Future work
- The second paper
- Dynamic adjustment of F and D
- The third paper
- A real prototype
- Experiments with real congestion control protocol
- Experiments with other types market-system, e.g.
auctioning system, or Lagrangian methods - Intelligent applications
- Adapt applications (media) using contextual
information
45Discussion
46Questions?
- Contact information
- johan.kristiansson_at_csee.ltu.se
- http//www.sm.luth.se/johank
- Visit this page for contact details
- Centre of Distance-spanning Technology (CDT)
- http//www.cdt.luth.se
- Division of Media Technology
- http///media.sm.luth.se