QoS Issues in Ad Hoc Wireless Networks

1
QoS Issues in Ad Hoc Wireless Networks

• Chakrabarti and Mishra
• IEEE Communications Magazine
• February 2001
• Presented by George Lee

2
ABSTRACT

• This article addresses some of the QoS (Quality
  of Service) issues for ad hoc networks.
• Focus on Qos Routing
• Complex and difficult
• dynamic nature of the network topology
• generally imprecise network state information
• The article concludes with some observations on
  the open areas for further investigation.

3
INTRODUCTION (1)

• Ad hoc wireless networks
• No fixed network infrastructures
• No centralized administration
• Self-creating
• Self-organizing
• Self-administering
• Fault-resilient
• Time-varying
• Multi-hop (nodes functioning as a routers)

4
INTRODUCTION (2)

• Challenges
• Effective routing
• How much and how often control information may be
  exchanged
• Medium (or channel) access
• Mobility management
• Instantaneous topology changes
• Power management ? Transmission range
• Limited bandwidth and hostile transmission
  characteristics of wireless channel
• Security
• QoS (quality of service)
• Delay and bandwidth management

5
INTRODUCTION (3)

• RFC 2386A Framework for QoS-based Routing in
  the Internet
• Agreement or guarantee (A set of measurable
  pre-specified service attributes)
• Transnetwork delay
• Delay variance (jitter)
• Available bandwidth
• Probability of packet loss
• and so on.

6
INTRODUCTION (4)

• Flow-based QoS
• A logical connection between a source and a
  destination
• Stateful router
• Resource reservation
• Link bandwidth
• Nodal buffers
• processing
• QoS Routing
• The process of choosing the routes to be used by
  the flow of packets of a logical connection in
  attaining the associated QoS guarantee.

7
AD HOC WIRELESS NETWORKS (1)

• Communication
• Single-hop(Direct communication)
• A B
• A D
• Multi-hop(Intermediate routers act as routers)
• A C
• A E

8
AD HOC WIRELESS NETWORKS (2)

• Beaconing
• The node broadcasts its address information (and
  may include its location information).
• Two nodes can establish direct communication to
  each other.
• Mobility of nodes ? route updates
• Ex. Path A-B-C but BC breaks? Reroute A-D-E-C

• As more nodes join or leave, the topology updates
  become numerous, complex, and more frequent, thus
  diminishing the network resources available for
  exchanging user information.

9
AD HOC WIRELESS NETWORKS (3)

• Combinatorially stable
• The topology changes occurs sufficiently slowly
  to allow successful propagation of all topology
  updates as necessary.
• The geographical distributions of mobile nodes do
  not change much relative to one another during
  the time interval of interest.
• Otherwise
• The last used route becomes unavailable.
• The QoS guarantees cannot be met.

10
AD HOC WIRELESS NETWORKS (4)

• Factors that affects QoS
• Connectivity
• Resource
• QoS-robust
• QoS guarantees are maintained regardless of the
  topology updates that may occur within the
  network.
• QoS-preserving
• QoS guarantees are maintained during the interval
  panning the end of a successful topology update
  until occurrence of the next topology change.
• QoS-roubust ? QoS-preservingQoS-preserving ?
  QoS-robust

?
11
QUALITY OF SERVICE (1)

• QoS guarantee
• To satisfy a set of predetermined service
  performance constraints
• End-to-end delay delay jitter
• Available bandwidth
• Probability of packet loss
• Two tasks for QoS routing
• To find a suitable path
• Resource reservation

• Example
• If the bandwidth guarantee from A to E is 3Mb/s
• Path A-B-C-E is selected

12
QUALITY OF SERVICE (2)

• Challenges of QoS routing
• Different service types have different objectives
  for delay, bandwidth, and packet loss.
• Determining candidate links is not simple.
• Route computation cannot be too long.
• More than one QoS constraints make the QoS
  routing problem NP-complete.
• Suboptimal algorithms such as sequential
  filtering are often used.
• A suboptimal path on a single primary metric is
  selected first.
• A subset of them are eliminated by optimizing
  over the secondary metric, and so on.
• A random selection is used if more than one
  choice remain after optimizing the last metric.
• The same route is used for all packets in the
  same flow.

13
QUALITY OF SERVICE (3)

• Resource reservation
• Once a route has been selected for a specific
  flow, the necessary resources (bandwidth, buffer
  space, etc.) must be reserved.
• These resources will not be available to other
  flows until the end of this flow.

14
QUALITY OF SERVICE (4)

• Minimization of routing updates
• Routing updates consume network bandwidth and
  router CPU capacity.
• Frequently changing routes increase the delay
  jitter.
• Difficult to attain in wireless networks
• Involuntary network state changes as nodes join
  or depart
• Traffic loads vary
• Link quality swing dramatically
• For real-time voice and video, minimizing the
  number of hops can keep delay and delay jitter
  under a certain bound.

15
QUALITY OF SERVICE (5)

• Network states
• Local state information
• Queuing delay and residual CPU capacity for the
  node
• Propagation delay, bandwidth, and cost metric for
  each outgoing links
• Maintained at each node
• Global state information
• The totality of local state information for all
  nodes
• Constructed by
• Broadcasting the local state of each node to
  every other node (link-state protocol)
• Exchanging distance-vector information among
  adjacent nodes only (distance-vector protocol)
• Maintained at each node
• Topology update
• The process of updating global state information

16
QUALITY OF SERVICE (6)

• Since topology updates throughout the network
  cannot happen instantaneously, the global state
  information may only be an approximation of the
  true current network state.
• Aggregated global state information
• Partitioning the network into a hierarchical
  cluster of some form
• Considering the state information associated with
  these clusters (partial representation of true
  global state)

17
QUALITY OF SERVICE (7)

• Three route finding techniques for QoS routing
• Source routing
• A feasible path is locally computed at the source
  node using the locally stored global.
• Then all other nodes along this feasible path are
  notified by the source of their preceding and
  successor nodes.
• Destination (distributed or hop-by-hop) routing
• The source as well as other nodes are involved in
  the path computation by identifying the adjacent
  router to which the node must forward the packet.
• Hierarchical routing
• Using the aggregated partial global state
  information to determine a feasible path
• Using source routing where the intermediate nodes
  are actually logical nodes representing a cluster
• Flooding is not an option for QoS routing except
  for broadcasting control packets (e.g. beaconing,
  route discovery)

18
QUALITY OF SERVICE (8)

• Different priories
• Control packets preemptive priority
• Data packets of different flows
• Authentication
• When a user requests QoS with certain priority
• Security
• Against unintended or deliberate attacks
• Flows making too many invalid requests
• Flows making inappropriate allocation of network
  resources
• To mimic or preempt network control functions
• Multicast

19
QOS ROUTING IN AD HOC NETWORKS (1)

• Unique identity
• Beacon
• Each node periodically broadcasts a beacon packet
  identifying it (and its pertinent QoS
  characteristics), thus allowing each node to
  learn of its adjacent neighbors (i.e. with which
  it can communicate directly).
• Two routing techniques
• One is based availability of only local state
  information
• The other assumes possibly inaccurate knowledge
  of global states.

20
QOS ROUTING IN AD HOC NETWORKS (2)

• Two distributed routing algorithms for QoS
  routing using the local state information
• Source-initiated routing
• Destination-initiated routing
• Both rely on the use of probe packets
• Sent by the source and intermediate routers
• To identify a feasible route with the desired QoS
  characteristics
• Using a form of flooding

21
QOS ROUTING IN AD HOC NETWORKS (3)

• The techniques based on imprecise knowledge of
  global states use ticket-based probing to
  identifying a feasible route.
• Each probe carries at least one ticket to control
  how many alternate paths to be searched, thus
  minimizing the routing overhead.
• The lower the likelihood of finding a route, the
  larger the number of tickets carried by the
  probe.

22
QOS ROUTING IN AD HOC NETWORKS (4)

• Detecting broken routes (using beaconing
  protocol), the node either
• Repairs the broken route
• Reroutes the flow on an alternate route with
  desired QoS
• If no alternate route segment is found, the node
  notifies the source.

• Figure 2
• s-A-B-C-D-d?s-A-B-E-C-D-d

23
QOS ROUTING IN AD HOC NETWORKS (5)

• When the source receives the notification of
  route unavailability, it seeks an alternate route
  with the same QoS characteristics and reroutes
  the flow to it.
• Reaffirming the existence of the QoS route
• Refresher packets are sent from the destination
  to the source periodically.
• If such a packet fails to arrive within a
  predetermined timeout interval, the QoS route is
  declared unavailable and the associated resource
  released.

24
QOS ROUTING IN AD HOC NETWORKS (6)

• Redundancy
• Highest level
• Multiple alternate routes with the same QoS
  guarantee
• Used simultaneously
• Preferably disjoint
• Middle level
• Routes and resources are reserved and ranked
• When the higher ranked route fails, the next
  ranked route is used.
• When not in use, the alternate route is used to
  carry best-effort packets.
• Lowest level
• Routes are identified, but resources are not
  reserved.
• When the primary route fails, the alternate route
  is checked and used.

25
CONCLUSION (1)

• Many of the underlying algorithmic problems are
  currently perceived as generally intractable
  (NP-complete).
• Lack of sufficiently accurate knowledge, both
  instantaneous and predictive, of the state of the
  networks (e.g. quality of the radio links, and
  available of routers and their resources).
• Guaranteeing QoS in such a network may be
  impossible if the nodes are too mobile.

26
CONCLUSION (2)

• The increased size of the ad hoc network causes
  increased computational load and difficulties in
  propagating network updates within given time
  bound.?Hierarchical ordered collection of
  subnetworks
• What is the proper size of each level?
• Is ordering always possible?

27
CONCLUSION (2)

• Coexistence of
• Best-effort service and QoS service (with
  different classes)
• Wireline networks and wireless networks
• Connectionless service and connect-oriented
  service
• ?
• Robustness
• Other challenges
• Network management
• Cost-effective implementation