A Locality Bounded Flat Hash Location Service

1
A Locality Bounded Flat Hash Location Service
HTL
Shuying Liang School of Software Shanghai Jiao
Tong University
Authors Ruonan Rao, Shuying Liang, Jinyuan You
2
Outline

• Motivation
• Related Works
• HTL (Our Contribution)
• Conclusion
• Q A

3
Introduction

• MANETs
• Challenge Scalable and Robust ROUTING PROTOCOLS
• Topologic Routing
• based on knowledge of the whole networks
  topological Information
• Geographic Routing - each message carries the
  position of the destination node (GPS..)

4
Geographic Routing

• Advantages of geo-routing
• No routing tables, efficient and scalable
• No setup time, each new node can participate
  immediately
• Fully reactive (on demand) routing
• Q1 How to deliver a packet to destination when
  the position of destination node is known?
• Answer forwarding strategies
• Greedy forwarding A node forwards a message to
  its neighbor closest to the destination
• Face routing To surround routing voids, a
  message may be routed along the border of the
  hole

5
Location Service

• Q2 How can a node determine the position of
  another node?
• Proactive distribution
• Each node broadcasts its position
• VERY expensive
• Home based location service
• Each node has an associated place
• (its home) where it stores its location
• The place is determined by the hash
• value of the nodes ID
• Any node can determine this place
• This is a GHT (Geographic Hash Table)

6
GHT

• Each node has a single location server
• Chosen at a random position
• Good load balancing
• Each node has to store location information only
  for a few nodes
• Big drawback It allows for a nearly arbitrary
  large stretch

7
Typical Related Work in the field of Location
Services

• GLS A Scalable Location Service for Geographic
  Ad Hoc Routing by Li et al. _at_ MobiCom 2000
• LLS LLS A Locality Aware Location Service for
  Mobile Ad Hoc Networks by Abraham et al. _at_
  DIALM-POMC 2004

8
Related Work (Excerpt) - GLS

• GLS Each node has several other nodes as
  location servers and also acts as a location
  server on behalf of some other nodes
• Lookup cost
• Size of the biggest surrounding
• square (level)
• Publish cost not bounded

9
GLS - worst case locality guarantees?

• Like GHS- NO
• there may be a source s and destination t that
  are arbitrarily close to each other, but the
  smallest square that contains both of them is
  arbitrarily large. As a consequence, even in
  average case networks, the cost of lookup does
  not have worst case bounds.

10
Related Work (Excerpt) - LLS

• LLS - a locality aware location service for
  mobile ad hoc networks.
• virtual grid of several layers with exponentially
  increasing size, server on level i does not store
  the location of the node, but only stores the
  location of the location servers of the next
  smaller level.
• Lookup cost O( d 2 )
• Publish cost (amortized)
• O(d log d)

11
LLS Locality Bounded?

• Bounds the lookup overhead to the square of the
  optimal path
• However, the cost of publish a location
  information and the cost of stored the location
  information is relatively high.

12
To Formally Evaluate the locality problem
-Locality Bounded

• Definition 1 (Locality Bounded) If the distance
  between the source node S and the destination
  node D is l, and the cost of query the location
  of D by S is at most f(l). We call such a
  location service is f-Locality Bounded.
• Can we design a geographic protocol that is
  locality bounded with a reasonable low storage
  and communication overhead?

13
HTL - A Locality Bounded Flat Location Service

• HTL
• Built on top of GPSR (Forwarding strategy)
• Lattice based physical space for query and update
  
• Publish Cost
• Storage Cost
• Lookup Cost
• Locality Bounded -gt O(d 2 )

14
HTL - GPSR

• Hybrid, two modes
• Greedy forwarding mode that moves packets
  progressively close to the destination at each
  hop
• Perimeter forwarding mode that forwards packets
  where greedy forwarding fails

15
HTL Algorithm - Lattice

• Mapping to physical space H(N.id) O(x,y)
  (h1(N.id), h2(N.id))
• Lattice Formation
• Lattice Circles draw concentric circles Ci using
  O(x,y) as the center of circles with the radius
  ri d i (I 1,2,), C0 is the center
  point
• Lattice Lines ,
  , ( j log2i )
• divide a circle into 2j2 with equal angle,
  anticlockwise
• Lattice Points
• Lattice Corners formed by
• 4 neighboring Lattice Points
• ----LATTICE formed Lattice Corners,
• arc of Lattice Circles and Lattice Lines

16
HTL - Location Servers Selection

• When publish node position information, location
  servers are selected along the path.
• Publish Process
• Lattice formed
• L (XN, Yn) -gt D H(N.id)
• Where L? Dist(L,D) d
• Can there be a path found? proved
• Server Coordination
• How to deliver packets to Server Coordination?
  GPSR

17
(No Transcript)
18
HTL - Lookup

• Node S wants to communicate with T, it firstly
  query location servers of T
• S finds D(H(T.id)) -gt construct lattice -gt locate
  S itself in which lattice
• Two modes
• Distent Mode forward packets to other Lattice
  Lines
• Non-distent Mode still in the current lattice
  lines

19
HTL - Lookup

• First, S forwards packets to two neighboring
  Lattice Points in a lattice corner in Distent
  Mode
• Then query will be forwarded to 2i Lattice points
• Location Server Points found, terminate
• If not found--------------?
• In step k, the query
• arrives H(D.id).If the
• location server knows
• where D is, then it reply
• the answer, tell S the
• location of D is unknown.

20
HTL Lookup (Example)
21
HTL - Location Update

• Nodes are allowed to move around, Update needed
• We define
• In the same Lattice
• No change in publish path
• But publish position update(P1, P3)
• In different lattices
• Path is needed to be changed
• Cancel former publish information,
• Republish position(P3, P4)
• Reconstruct Lattice (P5, P7)

22
HTL Analysis

• Parameters
• 
  
• 
  lt- (Based on lt 3 hops)
• 
  lt- (At least 1 hop)
• Publish Cost
• Storage Cost
• Lookup Cost Locality Bounded -gt O(d 2 )

23
HTL Analysis

• Result

24
Conclusion

• HTL Locality Bounded with the least overhead
  compared with existing routing protocols
• Discussion
• Is the O(d2) the best Locality Bounded can reach?
• HTL needs a lot computing capabilities, needed to
  be optimized

25
Questions / Comments

• Thank you!
• Questions / Comments?