Lookup Service for Peer-to-Peer Systems in Mobile Ad-hoc Networks

1
Lookup Service for Peer-to-Peer Systems in Mobile
Ad-hoc Networks

• M. Tech Project Presentation
• By,
• Kalpesh Patel
• 02329027
• Guided By,
• Dr. Sridhar Iyer,
• Dr. Krishna Paul.

2
Outline

• Introduction
• Motivation
• Problem Definition
• Related Work
• RINGS Lookup Service
• Analysis and Comparison
• Simulation Results
• Directions for Future Work and Conclusion

3
Introduction

• Peer-to-Peer Systems
• Nodes share files and services among other nodes
  in P2P network.
• Mostly focuse on wired network.
• Mobile Ad-hoc Networks
• MANET is an autonomous system for mobile hosts
  over wireless links.
• Multi-hop scenario.
• Mobility.

4
Motivation

• Why P2P in MANET?
• What happens if existing P2P protocols are
  deployed in MANET?
• Neighbours at application layer may not
  necessarily be neighbours at network layer.
• Problems with existing MANET protocols

5
Problem Definition

• The overall objective is to efficiently deploy
  P2P applications in Mobile Ad-hoc Networks.
• Specifically, need of a network layer protocol
  which optimizes query lookup service.

6
Related Work

• Mobile Peer-to-Peer Protocol (MPP) Recent work
  on P2P in mobile environment.
• Reference Rudiger Schollmeier, Ingo Gruber,
  Michael Finkenzeller, Protocol for Peer-to-Peer
  Networking in Mobile Environments.
• The problem is, MPP still relies on flooding at
  network layer.

7
RINGS Lookup Service

• Basic Protocol - Advertisement
• Network layer gets index from application layer.
• Node broadcasts its data index into the network.
• Every K-hop nodes keep index for a given node.
• This process is followed by every node in the
  network.
• Basic Protocol Query lookup
• Node forwards the query to its neighbors.
• Due to equal distribution of indices throughout
  the network, query is answered within K/2 hops.
• So, Upper bound for query lookup is K/2 hops.

8
RINGS Cont

• Basic protocol makes a set of imaginary circles
  for a given source node as shown in Fig.1.
• Fig. 2 shows an example of 2-hop (K2) scenario.

9
RINGS Cont

• Route Maintenance
• Advertise packet maintains route to the source
  node.
• This is accomplished by maintaining reverse route
  on the path.
• A node, which stores the index, keeps this route
  along with other information.
• If a query comes to this node, it responds with
  query result as well as route to the source.
• An optimal route.

10
RINGS Cont

• Mobility
• Mobility has none or little impact on the
  performance of the protocol.
• In average, number of neighbors per node in
  mobile network can be taken as log(N).
• So, even if as much as log(N) -1 nodes moves
  away from the network, node gets at least one
  node for query result.
• Thus, a node gets at least one node which can
  satisfy its query even if L number of nodes from
  nearest circle leave the network, where L is,
• Where K
  2, 4, 6,

11
RINGS Cont

• Optimization
• Data Updates
• Updates are sent to only those nodes which have
  stored indices for the source node.
• Node Joining
• Whenever a new node joins the network, it gets
  some share of indices from its neighbors.

12
RINGS Cont

• Node Deletion
• Neighbors of leaving node search for an
  alternative route.
• After getting alternative route, indices are
  reallocated according to index-hop (K) value.

13
Analysis

• Goal
• Count the actual query lookup cost for P2P
  protocol like Chord in MANET.
• The average distance d between any two nodes at
  the network layer can be taken as,
• Now, query lookup cost for Chord in application
  layer is log(N).
• Actual cost App. Layer cost Network layer
  cost.
• So, effective lookup cost Lchord for chord
  protocol is,

14
Analysis (Cont)

• For RINGS,
• Query lookup cost at first hop is,
• In turn, each of these log(N) nodes forwards
  query to its neighbors.
• Continuing this way, query lookup cost at the
  nearest K-hop is,

15
Simulation Results

• An important measure is, how average number of
  remote indices stored per node increases as the
  value of K increases.
• Result indicates that, as value of K increases,
  average number of remote indices per node
  increases very fast.
• Value K 4 balances both the metrics.

16
Simulation Results (Cont)

• Graph shows comparison between query lookup costs
  Lchord and LRINGS
• For the value of K 4, RINGS performs better
  than Chord in terms of query lookup cost.
• For the value of K 6, as the number of nodes
  increases beyond a certain number, RINGS performs
  better than Chord.

17
Comparison

• Why RINGS?
• Most recent work still relies on network layer
  flooding for query lookup.
• Most adopted MANET protocols like AODV, DSR uses
  broadcast technique, results in flooding the
  whole network every-time when a search for a node
  requires.
• P2P protocols
• Gnutella uses broadcast.
• Chord assigns every node a virtual identifier
  independent of actual locations, which is not
  preferable at network layer.

18
Directions for Future Work

• RINGS assumes enough storage space available at
  the node. There is a scope of improvement to
  reduce this overhead.
• Effective mobility handling.
• Practical applications.

19
Conclusion

• There is a need to combine P2P applications and
  Ad-hoc systems to share data among mobile users.
• Current P2P protocols fail to perform well with
  this combination.
• An initial step to this problem is RINGS, which
  focuses on query lookup cost optimization.

20
Thank You