Distributed Allocation of Location Information DALI: Location Service for Mobile Networks

1
Distributed Allocation of Location Information
(DALI) Location Service for Mobile Networks

• Jayasri Akella, Vijay Subramanian, Xiaobo Long
• ECSE 6962 Final Project Presentation

2
What is a Location Service?

• A location service is used by the sender of a
  packet to determine the position of the
  destination. MM01
• Need for Location Services
• For Position-based routing, a node needs to know
  the current position of the destination.
• Centralized solution does not scale and presents
  a single point of failure.
• A distributed and scalable solution is needed to
  aid location-based routing such as GPRS, GEAR
  etc.

Source A
Query and Response
Location Server C
3
Problem Statement

• Given a name, find a nodes position!
• For an ad hoc network of arbitrary size, provide
  an accurate, scalable and efficient location
  service.
• We also wish to minimize the following overheads
• Number of control messages.
• Number of queries and replies.
• We also want to maximize the success rate of the
  queries.
• Accuracy The reply to the query must be
  up-to-date and must not return the position of a
  node that is outdated,

4
Assumptions

• Each node knows its own position.
• Solutions exist for localized positioning.
• Each node knows the boundary of the network.
• All the nodes share a hash function discussed
  later.
• All the nodes know the partitioning of the
  network.
• All regions have at least one node in all regions
  at all times.
• Probability of an empty region can be found based
  on the mobility model.

5
Basic Concepts

• Network is divided into K regions each of which
  has 1 or more nodes and 1 cluster-head.
• Home Area Region to which a node belongs.
• Queries about a nodes current location will be
  resolved in the home area of the node.
• Nodes present in the home area will share
  information about the current locations of the
  node.
• A home area may be shared among several nodes but
  a node has only 1 home area.
• Each home area has a area-head that is near the
  center of the home area and receives queries.

6
Our Approach DALI (Distributed Allocation of
Location Information)
Q1
Source A
Destination B
7
Problem Statement
Source A
8
Overview of DALI

• Initial Bootstrapping phase.
• Three steps
• Finding the home region of the destination.
• Identifying location Servers for the destination.
• Querying location service and obtaining the
  current location.
• Y wants to communicate with X
• Y hashes node X's ID to a region
• If X is at Ys home region, internal hashing can
  be used to get its location.
• If X is not at Ys home region, global hashing
  can be used to get the location servers.
• This minimizes traffic and delay while increasing
  efficiency.

9
Detailed Description

• A nodes home region can be inferred from the
  nodes id or name.
• A hashing function can be used for this purpose.
• Each node periodically updates its location
  servers with its current location.
• All the nodes in a home region act as location
  servers of the nodes with in that region.
• divide network into a few geographic regions
• can better direct a location query and achieve a
  higher probability of success.
• At all times, there will be at least one node
  present in a region to carry the pointers for all
  the nodes
• We will find the probability of not having any
  node in the region given a mobility model
• This will also leads to the optimal splitting of
  regions
• Or shall we do dynamic splitting?

10
Tackling Mobility

• When a node leaves its current region and moves
  into a different region, it changes its home
  region?
• Queries will be sent to the new region
  henceforth.
• The old home region will redirect any queries it
  receives.
• If old home region is empty, Directed queries to
  cluster heads can be performed.

11
Distributed Hash Function

• Chord
• Maps a key onto a node
• provides load balance
• Fully distributed
• no node is more important than any other
• Scalability
• grows as O(logn), where n is number of nodes
• Availability
• adapts to newly joined nodes as well as nodes
  failures

12
Consistent Hashing

• Chord uses consistent hashing
• to provide fast distributed hash function
• M-bit identifier
• each node and key is assigned an M-bit identifier
• Node identifier chosen by hashing nodes ID
• Using bash hash function such as SHA-1
• Key identifier chosen by hashing the key
• M must be large enough
• Make the probability of two nodes or keys hashing
  to the same identifier negligible

13
Consistent Hashing (cont.)

• Assigns keys to nodes
• Identifiers are ordered in an identifier circle
  modulo 2M
• Key K is assigned to the first node whose
  identifier equal or follows k in the identifier
  space
• The node is called successor node of key k,
  denoted by successor(k)
• Here key is nodes name
• Use consistent hashing to assign keys to 2M
  location servers
• Given a nodes name? hashing to its location
  servers within its region

14
Example of how consistent hashing assigns keys to
nodesM3three nodes with identifiers
0,1,3238 keys identifiers 0,1,28key1success
or of key 1 is 1, so key 1 is assigned to node
1key2-3successor of key 2-3 is 3, so key 2-3
is assigned to node 3key 4-7successor of key
4-7 and key 0 is 0, so key 4-7 and key 0 is
assigned to node 0
15
Consistent Hashing (cont.)

• When nodes enter and leave network
• Aim minimum disruption
• Theorem 1. For any set of of N nodes and K keys,
  with high probability
• Each node is responsible for at most (1e)K/N
  keys, where eO(logN)
• When an N1th node joins or leaves the network,
  responsibility for O(K/N) keys changes hands (and
  only to or from the joining or leaving node)
• Example Nodes join
• Certain keys previously assigned to ns successor
  now assigned to n

16
Scalable Key Location

• How to implement queries for a given key
  identifier
• Method 1
• Pass the key identifier around the circle
• until first encounter a node that succeeds the
  identifier
• Not efficient
• Method 2 Finger table

17
Finger Table

• M the number of bits in key/node identifier
• Each node maintains a routing table with (at
  most) M entries
• The ith entry contains the identity of the first
  node s that succeed n by at least 2i-1 on the
  identifier circle
• Note
• The first finger of n is its immediate successor
  on the circle
• Each node stores information about only a small
  number of nodes
• Each node knows more nodes closely following it
  on the identifier circle than about nodes farther
  away
• A nodes finger table doesnt contain enough
  information to determine the successor of
  arbitrary key

18
Finger Table (cont.)

• What happens when a node n does not know the
  successor of a key k?
• If n can find a node whose ID is closer than its
  own to k, that node will know more about the
  identifier circle in the region of k than n does.
  
• Thus N searches its finger table for the node j
  whose ID most immediately precedes k, and asks j
  for the node it knows whose ID is closest to k.
• By repeating this process, n learns about nodes
  with IDs closer and closer to k.

19
Selecting and Querying Location Servers

• Our design RLS (Region based location service)
• Reference GLS (Grid location service)
• RLS provides distributed location service
• By replicating the knowledge of a nodes current
  location at a small subset of the network
• This set of nodes is referred to as the nodes
  location servers
• Node A hoping to contact node B can query one of
  a number of other nodes that know Bs location

20
Two Hash Tables

• Internal hashing
• Load balance for location servers within a region
• Global hashing
• Load balance for location servers outside a
  region
• Each node maintains two finger tables
• One for internal hashing
• One for global hashing

21
Hierarchy Regions-for better scalability
22
Related Work

• Decentralized Location Services
• DREAM Distance Routing Effect Algorithm for
  Mobility.
• This is an all-for-all approach.

23
References

• MM01A Survey on Position-Based Routing in
  Mobile Ad-Hoc Networks (2001) Martin Mauve, Jörg
  Widmer, Hannes Hartenstein.
• IS99A Routing Strategy and Quorum Based
  Location Update Scheme for Ad Hoc Wireless
  Networks (1999) Ivan Stojmenovic, Bosko Vukojevic
• IB A Quorum-Based Dynamic Location Management
  Method for Mobile Computings Ihn-Han Bae
• IS Home agent based location update and
  destination searchschemes in ad hoc ireless
  networks. Ivan Stojmenovic
• Robert Morris?

24
Summary