Folie 1

1
Cross-Layer Integration in Ad-Hoc Networks with
Enhanced Best-Effort Quality-of-Service Guarantees
Christian Webel University of Kaiserslauternwebel
_at_informatik.uni-kl.de
WTC 2006, Budapest 01.05.2006
2
Motivation

• distributed applications require a specific
  network Quality of Service
• connectivity, throughput, delay, QoS guarantee
• problem
• varying channel conditions
• varying connectivity
• interfering nodes
• ? deterministic or permanent statistical
  guarantees are beyond reach
• ? new QoS guarantees and new QoS protocols are
  needed
• ? enhanced best-effort

3
Case study Airship System

• one airship
• one remote pilot
• one or more video projectionfacilities
• hardware
• airship with helium
• battery-poweredembedded computer
• wireless LAN (IEEE 802.11b)
• miscellaneous hardware to steer the airship
• camera for surveillance during the flight

4
Airship System Architecture

• Application Layer
• distributed flight control system
• video transmission system

• Middleware Layer
• mechanisms to steer the airshipand to adjust and
  transmit the video stream
• Basic Service Layer
• controls access to the communication medium
• QoS MAC layer with slot-based bandwidth
  reservation on top of IEEE 802.11b (WLAN)

5
Enhanced Best-Effort

• service differentiation on application level
• proactive
• distributed admission control and bandwidth
  reservation algorithm
• periods of statistical guarantees
• reactive
• priority-based resource management and scaling
• QoS degradation
• QoS cross-layer integration
• ? better than best-effort
• application domain
• ambient intelligent systems with scarce and
  varying resources

6
QoS Specification (1)

• part of the traffic contract between service user
  and provider
• performance specification
• QoS management policies
• QoS-guarantee (e.g. enhanced best-effort with
  priority 3)
• concrete parameters vary with the kind of service
  and the viewpoint

7
QoS Specification (2)

• e.g. video transmission
• user point of view QoS types denoting the
  requested user QoS, e.g., panorama ? video
  transmission serving surveillance purposes.
• application level image resolution (in pixels),
  number of images per second, image quality
  (minimum and optimum)
• network QoS on middleware level number of data
  frames per period. Period depends on the number
  of images per second.
• network QoS on basic service level bytes per
  second. Delay and jitter may be defined.

8
Overall System Architecture
Middleware
Scaling
AppMapping

• QoS cross-layer integration
• feedback cycle
• controls the used bandwidth
• priority based

MwMapping
ResourceManager
BasicService
PacketBuffer
Monitor
Feedback Cycle Data Flow Control Flow
GlobalScheduling
9
QoS Provision

• negotiation of a traffic contract
• available resources are checked and allocated
  according to the specification
• example QoS Mapping (QoS cross-layer
  integration)
• different layers use different kinds of QoS
  specifications
• a mapping between those layers is required
• e.g., mapping application layer ? middleware layer

10
QoS Control

• short-term mechanisms applied by the service
  provider to fulfill his traffic contracts.
• reactive mechanisms
• example Traffic Shaping
• is applied if an application violates the traffic
  contract
• traffic is reduced by discarding or delaying
  packets
• flight control system
• modified leaky bucket algorithm with queue
  length of one

11
QoS Management

• medium- and long-term mechanisms to support the
  adaptation of QoS as well as the improvement of
  network performance.
• reactive and proactive mechanisms
• example priority based QoS Scaling
• closed-loop control calculates the available
  quality of service
• based on the network load
• available quality of service is split up for the
  different applications according to their
  priorities
• application scaling is done in accordance with
  the minimum and optimum parameters of the current
  quality of service class

12
Model-Driven Development

• completely specified using the ITU-Ts
  Specification and Description Language (SDL)
• SDL environment framework for platform
  independency
• automatic generation of code from the design
  specification
• no handcrafted coding was done

13
Performance Simulation (1)

• nsSDL
• extension of ns-2
• network simulator for SDL systems
• Goal evaluation of the adaptive behaviour of the
  airship system
• additional load is placed on the network.
• fixed share of 65 of the overall usable
  bandwidth
• to cope with variable channel quality
• to leave some bandwidth for regular best-effort
  communication
• t30 sec a new application reserves bandwidth
  with a higher priority than the video
  transmission
• t45 sec the application terminating its
  transmission

14
Performance Simulation (2)
monitoring
video stream
temporary sender
controller

• tsim30 sec, 45 sec adaptation of the video
  stream to the new resource situation
• high priority control data for the airship was
  not affected at any time

15
Conclusions

• Enhanced Best-Effort
• new type of QoS guarantee
• realized on top of WLAN
• Ad-hoc networks
• scarce and varying resources
• adaptive QoS mechanisms
• QoS cross layer integration
• simulations to assess the performance

16
Demo Airship in Operation