Information distribution by mobile agents in circular path

1
Information distribution by mobile agents in
circular path

• Keita Fujii

2
Idea

• In animals body, there is a network of blood
  vessel that carries energies throughout the body
• Blood also carries some information such as
  hormone and blood sugar level
• Can we apply the concept of blood vessel into
  Bionet?

3
Basic concept

• Several CEs (say, info CE) are traveling around
  the network along a circular path
• These CEs have information of normal CEs
• Normal CEs can either send their advertisements
  or send queries to those info CEs
• CE can communicate only with others in a certain
  range

4
Basic concept (contd)

• There may exist multiple circular paths
• Several path may integrate as one
• The path may change according to the demand and
  situation of the network
• Use pheromone to change the path
• Each info CE may take different path
• Just like each red blood cell may take different
  vessels (but comes back to the same vessel later)
• Info CE may reproduce to process higher demand
• Info CE may die if theres no enough demand

5
Benefits

• Not centralized
• No single point of failure
• Can adapt to the network
• Not distributed (hierarchical) database
• Each info CE may contain different information
  (not like synchronized distributed database
  approach)
• Not P2P broadcast
• Less traffic More efficient
• Matches the concept of Bionet

6
Research Issue

• How to set up a circular path?
• How to guarantee that information is advertised
  globally?
• How to detect the demand of advertisement or
  query?
• How info CE can evolve?

7
Schema for circular path creation (1) use graph

• Info CE can obtain topology information of
  Platforms
• Info CE calculates a circular path from the
  topology graph
• Info CE may concern some factors like demand,
  bandwidth, memory space etc
• Problems
• It is almost impossible to obtain a whole
  topology information of ad-hoc P2P network. It is
  likely that only local topology information is
  available. If so, the resulting path may not be
  adapted to the whole network
• Traveling Salesman Problem NP Complete

8
Schema for circular path creation (2) dynamic
path adaptation

• Start from roundtrip on the single path
• CE changes the path if
• Demand is detected
• Probabilistically
• Next node/link is inaccessible
• Path is extended in either of the following ways
• Go-and-back
• Choose a new node, migrate to there, then come
  back
• Replace one
• Replace the next destination with a new node
• Skip one
• Skip the next destination and migrate the 2nd
  next one

9

• New node is chosen based on
• Demand
• Resource (memory size, bandwidth)
• Randomly
• Path is optimized at some timing
• At some interval
• When replicated
• When reproduced

10
(No Transcript)
11
Demand detection mechanism (1) Wave Scheme

• When a normal CE advertises its information or
  sends a query, Platform broadcasts it locally (
  small TTL)
• If the advertisement/query isnt captured by info
  CE in a certain time period, Platform broadcasts
  a negative wave packet globally ( big TTL)
• Info CE emits a positive wave packet
• Counteract with negative wave
• Info CE detects negative wave and move towards
  its source

12

• Wave packet can be aggregated to reduce traffic
  consumption
• Wave strengths are summed up before being
  forwarded to next Platform
• Q Should wave strength be weaker as traveling
  farer?
• If it is, then the farer the less likely to be
  found?
• If not, when should wave packet stop?

13
Demand Detection Mechanism(2) Differential
Scheme (?)

• Each platform has its demand point
• Higher point means more demand exists
• Each platform updates its demand point
  periodically
• The demand point of a platform is the average of
  the demand points of the platform around it
  number of uncaptured advertisement/query on the
  platform
• Info CEs try to minimize the average demand point
  of the network by visiting platforms with high
  demand points

14
Information Exchange between info CEs

• Do info CEs exchange information about
• Path?
• Demand?
• Advertisements?
• I think they should, but we must examine whether
  it would create clustering of info CE or consumes
  inefficient amount of resource

15

• If info CEs are allowed to exchange advertisement
  information, it is possible to create a chain of
  circular path to propagate advertisements

16
Evolution of info CE

• Parameter of info CE
• Path information
• Path change policy
• Migration speed
• Interaction policy between info CE
• Capacity of advertisements
• These parameters can be changed at the evolution
  phase to achieve more efficiency and adaptability

17
Multiple types of info CE

• Introduce several types of info CEs with
  different characteristics
• Path is global/local
• Migration speed is faster/slower
• Maximum length of period it can hold
  advertisement is short/long
• User / normal CE can choose which type of info CE
  he may want to use
• If he wants to advertise real-time information
  locally, he may want to choose local/faster
  infoCE
• If he wants to advertise stable information
  globally, he may want to choose
  global/slower/long period infoCE

18
Simulation Plan

• Three simulations are possible (necessary)
• Comparison between this model, centralized
  (server) model, hierarchical model (like DNS) and
  P2P broadcast model
• In static environment
• In dynamic environment (ad-hoc network, mobile
  object)
• Comparison of different scheme for path
  adaptation/demand detection
• Evaluation of evolution scheme for info CE

19

• Target measurement parameters
• Successful query ratio
• Success query / total query
• Response time
• Time period from query sent to response back
• Necessary Traffic
• of bytes (or packets) consumed
• Freshness of data
• Time period from advertised to available to
  everybody
• Robustness against node/link trouble