AntSearch: An Ant Search Algorithm in Unstructured

Peer-to-Peer Networks

Apoorva JOSHI

Somanshu DHINGRA

IST-NET2
INSA Lyon

What we will discuss?

Introduction

4. Advantages / Disadvantages

3. Antsearch Algorithm

2. Need for a new algorithm

5. Q and As

Introduction

Popularity of unstructured P2P

Scalability
Decentralization
Reliabilty
Heterogenity

Problems

Random Flooding:

Large network traffic
Poor search efficiency
All files may not be returned (constrained TTL)

Improvement

Dynamic Querying :

Based on the popularity of file

Computes a TTL dynamically for each neighbour

BUT....

Freeriders

The root of all problems

~ 96% of the peers are free riders

Make the network congested

Slows down searching

Antsearch

Dynamic Querying

Overcoming Freeriders

Antsearch

Antsearch: Algorithm

Overcoming Freeriders

Pheromone Table:

Probability of being a Freerider

Text

Alpha defines the weight for each attribute.
Enables even a new node to be searched.

When is the value of pheromone table updated?

1) When a peer joins into the network

2) When a peer leaves the network

3) When a query is sent to the peers.

=>Low maintenance cost!

Antsearch: Algorithm

Probe Phase : Statistics about total number of searched files and corresponding number of searched peers is collected and a probe table is created using this data.

Antsearch: Algorithm

2. Flooding Phase: In order to search 'N' files, TTL is computed for each k. The least value of k is chosen, which satisfies the condition TTL< TTL_max.

Antsearch: Algorithm

TTl(k,nk,D,N)

Antsearch: Algorithm

Title Text

Antsearch vs DQ

Reduces traffic load by 50% compared to DQ.
Low maintenance cost.

Advantages

Load imbalance issues
For a rare file, it behaves like a TTL constrained random flooding.