Virtual Ring Routing Network Routing Inspired by DHTs

1
Virtual Ring Routing Network Routing Inspired by
DHTs

• Matt Caesar, Miguel Castro, Ed Nightingale,
• Greg O'Shea, Ant Rowstron
• Microsoft Research

2
Previous work on wireless routing

• flooding-based protocols
• proactive protocols flood on topology changes
• reactive protocols flood to discover routes
• poor performance when routes fail
• coordinate-based and hierarchical protocols
• nodes have location-dependent addresses
• DHT-like structure to translate ids to addresses
• expensive to deal with greedy forwarding failures
  

3
Virtual Ring Routing (VRR)

• unique design inspired by DHTs
• implements point-to-point routing
• implements a distributed hash table
• but runs just above the link level
• can it provide better wireless routing?
• no flooding
• no location-dependent addresses
• robust performance

4
Talk overview

• introduction
• VRR
• performance
• conclusion

5
VRR the virtual ring
topology-independent node identifiers
910
90E
8F6
8F0
8E2
each node maintains a virtual neighbor set (vset)
nodes organized into virtual ring by increasing
identifier value
6
VRR routing paths
8F6
8F6
physical network topology

• nodes only maintain paths to virtual neighbors
• vset-paths are typically multi-hop
• vset-paths maintained proactively

7
VRR forwarding table
14A
F42
endpointA
endpointB
nextA
nextB
pathId
31
8F6
90E
ME
F42
910
8F6
10E
ME
10
8F6
14A
140
F42
10E
2
F42
ME
F42
8F6
FF
10E
forwarding table for node 8F6

• vset-paths recorded in forwarding tables along
  path
• forwarding table contains
• vset-paths between node and vset members
• vset-paths between other nodes that go through
  node
• paths to physical neighbors

140
8
VRR forwarding

• forward message destined to x by
• picking endpoint e numerically closest to x
• forwarding message to next hop towards e
• deliver message to node with id closest to x
• how does this work ?
• can find x because nodes are connected in a ring
• low stretch because of additional forwarding
  state
• many alternate paths to route around failures

9
VRR example routing
physical network Topology
10
VRR example routing
there may be some stretch
physical network Topology
11
Node joining
broadcast hellos to find physical neighbors
send setup request to 16E through proxy
164
8F6
19A
Network Topology
16E
12
Node joining
16E sends setup requests to nodes in received
vset
164 sends setup to 16E with its vset
16E adds node to vset when it receives setup
164
171
8F6
19A
Network Topology
16E
13
Simulation experiments in ns-2

• ran experiments with 802.11b MAC
• varied network size, mobility, session lifetime
• compared with DSDV, DSR, and AODV
• VRR performed well in all experiments
• high delivery ratios even with fast movement
• significantly lower delays with route instability

14
Delivery ratio fast movement
15
Delay fast movement
16
Sensor network

• sensor network testbed
• 67 mica2dot motes in UCB building
• comparison with BVR
• delivery ratio with mote failures

17
Sensor network mote failures
18
Wireless office testbed

• 30 machines running windows
• communicate using 802.11a
• throughput comparison with LQSR using ttcp

19
Wireless office testbed throughput
20
Conclusion

• overlay routing at the network layer
• VRR implements wireless routing
• no flooding
• single location-independent identifier per node
• VRR implements a DHT
• strong consistency
• no additional overhead

21
Code is available for download

• http//research.microsoft.com/vrr

22
Virtual ring consistency

• vset piggybacked on messages
• nodes check received vsets
• they send setup requests to candidates
• candidates may reply with setup messages
• nodes update their vset when they receive setups
• nodes send setup requests
• when they join
• when they receive a teardown

23
DHTs and wireless ad hoc networks

• distributed hash tables (DHTs)
• map object keys to nodes
• provide efficient lookups of objects given keys
• are self-organizing, fault-tolerant, and scalable
  
• require no infrastructure
• perfect in wireless ad hoc networks
• sensor networks, mesh networks, emergency and
  disaster relief networks

24
How do we build DHTs on wireless ?

• DHTs implemented as overlays
• but layering on wireless routing is inefficient
• poor interaction between DHT and routing layers
• problems with wireless routing protocols

25
Problems with wireless routing

• proactive algorithms
• maintain routes between all pairs of nodes
• flood information about topology changes
• on demand algorithms
• find routes on demand by flooding
• geographical and landmark routing
• route using coordinates
• must translate between identifier and coordinates
• high overhead with movement and large stretch

26
VRR interesting properties

• stretch empirically small
• 40 with 200 nodes with vset 4
• many alternate paths to route around failures
• randomization balances forwarding load

27
Maintenance

• no flooding
• all messages routed using algorithm described
• five message types
• hellos maintain physical neighbor set
• setups update forwarding table state along path
• setup requests ask another node to send setup
• setup replies indicate refusal to setup path
• teardowns remove forwarding table state

28
Handling failures

• routing state is hard
• no end-to-end heartbeats
• failures detected on missing acks or hellos
• local repair attempted first
• otherwise, teardowns sent along all affected
  paths
• two techniques to ensure consistency
• symmetric failure detection and acks on teardowns
• if x marks y faulty, y is guaranteed to mark x
  faulty
• lightweight optimistic transactions

29
A link failure example
Endpoint B
Network Topology
Endpoint A

• repair is truly local
• only involves nodes near failed link or node
• no end-to-end path metrics
• repair aborted if local consistency checks fail

30
Virtual ring consistency

• vset piggybacked on messages
• nodes check received vsets
• they send setup requests to candidates
• candidates may reply with setup messages
• nodes update their vset when they receive setups
• nodes send setup requests
• when they join
• when they receive a teardown

31
Sensor network increasing load
32
Delivery ratio no movement
33
Delay no movement
34
Delivery ratio short sessions, fast movement
35
DHTs and wireless ad hoc networks

• distributed hash tables (DHTs)
• map object keys to nodes
• provide efficient object lookups
• are self-organizing, fault-tolerant, and scalable
  
• require no infrastructure
• perfect in wireless ad hoc networks
• sensor networks, mesh networks, emergency and
  disaster relief networks