Problem Statement for OSPF Extensions for Mobile Ad Hoc Routing

1
Problem Statement for OSPF Extensions for Mobile
Ad Hoc Routing

• Fred Baker
• Cisco Systems
• draft-baker-manet-ospf-problem-statement-00.txt

2
Interesting drafts

• Things we think are close
• draft-mirtorabi-ospfv3-af-alt-00.txt
• 3 more drafts coming
• Related Work
• draft-ogier-manet-ospf-extension-00.txt
• draft-spagnolo-manet-ospf-wireless-interface-00.tx
  t

3
Fundamental Approach

• Current target
• US Army WIN-T Network
• Design Guidelines
• Add a radio interface to OSPF without perturbing
  existing protocol or capabilities
• Using that within a combined radiowired network
• Collaboration between
• Cisco folks
• Inria (OLSR) folks
• Boeing Research
• Naval Research Labs

4
Security considerations

• Mobile wireless networks prone to physical
  security threats
• Inter-router authentication needed prior to the
  exchange of network control information.

5
Network design requirements

• Manet networks are not often isolated networks
• Manet networks often consist of vehicles
  containing racks with equipment and an internal
  LAN, or vehicles or robots accompanying groups of
  people connected to them via wireless LAN
  technologies and interconnected with other such
  groups.
• Often interspersed with field offices
• Connected to a larger internet via standard WAN
  technologies such as satellite links, optical
  fiber, or virtual and circuit switched networks.

6
Combining manet and traditional wired networks

• Adaptation of both OSPFv2 and OSPFv3 has been
  considered

7
The big picture
Service backbone Such as satcom to Fixed assets
Manet cloud
Truck addressable Within regional
rapid Deployment network
Unit-specific Fixed assets
Truck addressable Within fixed base Network
(802.11)
NIPRNET etc
Base network
8
Router-to-radio connection
System
Next-generation 3200
Radio
Different rates depending on SNR/BER and radio
waveform of connection (potentially many
waveforms in same chassis)
9
Key considerations in traffic management

• Router-radio interface has what amounts to
  virtual circuits per neighbor
• Shallow txring in radio per routing neighbor
• Separate MQC QoS data structures per routing
  neighbor
• Each routing neighbor has a different rate,
  managed by the radio
• Therefore each neighbor must have its own
  controlled rate protocol between router and radio

System
PPPOE with extensions
Next-generation 3200
JTRS Radio
10
Mobile ad hoc network issues

• Often bandwidth-constrained
• Often power-constrained
• Limited security
• Unnecessary transmission is something to be
  avoided.
• Security threats need to be actively addressed,
• Transmissions and computation loads need to be
  minimized.

11
Ad hoc networks

• IEEE 802.11?
• Choose layer 3 routing over 802.1 bridging?
• Network management concerns,
• Scaling concerns,
• Administrative boundaries
• No obvious way to aggregate neighbor sets in the
  way one does for lans.

12
Relay traffic minimization

• Optimizations such as DR election will not
  perform correctly in MANET networks
• How much information needs to be flooded
  throughout the network?
• Ideal solution each node will receive a
  particular lsa exactly once
• It would be nice if only and exactly one router
  relayed lsas.
• It would be nice to minimize the transmission of
  lsa acknowledgements

13
Standard OSPF Flooding
14
Overlapping Relays
15
Addressing power issues

• Units that rely on batteries may set a policy
  that they are willing to relay messages on their
  radio interfaces, but not very willing
• May also have LAN interfaces and be willing to
  relay traffic from radio interfaces on which they
  are not exchanging even keep-alive traffic to
  non-radio neighbors at times that their radio
  interfaces are incommunicado.

16
Scaling for rapid movement

• Network should not thrash
• If neighbor relationships are changing rapidly,
  the time invested to become FULL with a peer as
  defined in OSPF must be minimized,
• Efficiently scope the propagation of topology
  information such that rapid mobility or change in
  one part of the network need not be disseminated
  throughout the network.

17
IPv4 and IPv6 routing

• OSPF has historically chosen a "ships in the
  night" model
• In mobile ad hoc networks, however, running two
  protocols in parallel has a heavy impact when the
  network changes
• In mobile ad hoc networks carrying both ipv4 and
  ipv6 traffic, it would be good for a single
  routing protocol to calculate one set of routes
  and decorate those routes with both ipv4 and ipv6
  prefixes.

18
Reliable vs unreliable flooding

• Problem with unreliable flooding the assumptions
  it makes.
• Frequent unreliable flooding limited scalability
• Consequences of losing even a single LSA can be
  great.
• Exponential explosions of transmissions, such as
  can be prohibitively chatty

19
Area hierarchy

• No present manet protocol provides a counterpart
  to OSPF administrative areas.
• Some provide dynamic clustering
• Not obvious that the hierarchical area model fits
  proposed manet deployments.
• Model permitting summary lsas exchange at any
  boundary preferable

20
Problem Statement for OSPF Extensions for Mobile
Ad Hoc Routing

• Fred Baker
• Cisco Systems
• draft-baker-manet-OSPF-problem-statement-00