Bio-Net: A Biologically Inspired Architecture for Adaptive Network Applications

1
Bio-Net A Biologically Inspired Architecture
for Adaptive Network Applications

• Tatsuya Suda
• Information and Computer Science
• University of California, Irvine
• suda_at_ics.uci.edu

2
Outline

• Motivation and Biological Concepts in Bio-Net
• Emergent Behavior
• Evolution and Adaptation
• Peer to Peer Discovery in Bio Net
• Current Status of Other Subprojects
• Adaptation and evolution simulations
• Platform design
• Service emergence
• Conclusions

3
Motivation

• NGI applications need to be
• scalable, adaptable, survivable/available
• Observation
• large scale biological systems have desirable
  features
• So, apply biological concepts/mechanisms

4
Emergent Behavior

• Biological systems
• (useful) group behavior emerges from autonomous
  interaction of individuals with simple behaviors

5
Emergent Behavior in Bio-Net

• individuals cyber-entities (CEs)
• abstraction of system components
• users, resources, service components (e.g.,
  flight reservation service component)
• autonomous with simple behaviors
• replication, reproduction, migration, death, etc.

6

• CE behavior energy exchange
• gain energy from a CE (e.g., a user) in exchange
  for performing a service
• expend energy to receive service from other CEs
  (e.g., to use network/computing resources)
• energy as a natural selection mechanism
• death from energy starvation
• tendency to replicate/reproduce from energy
  abundance

7

• CE behavior relationship establishment
• a CE knows something (e.g., name, address,
  service type) about another CE
• Relationship strength (useful strong)

8

• relationship to group CEs collectively providing
  an application
• application constructed from a collection of CEs
• e.g., a web server (application) from a
  collection of web pages (CEs)

9

• relationship for application emergence
• addition/deletion of new CEs to an application

10

• relationship for application emergence
• addition/deletion of new CEs to an application

11

• relationship for application emergence
• addition/deletion of new CEs to an application

12
Evolution and Adaptation in Bio-Net

• Diversity and Natural Selection in Bio-Net
• CEs evolve/adapt through
• Diversity
• A CE behavior implemented by a number of policies
• By human designers
• By mutation/crossover in CE replication/reproducti
  on
• Natural selection (using energy)
• death from energy starvation
• tendency to replicate/reproduce from energy
  abundance

13
Vision

• No central or coordinating entity exists.
• A large number of CEs (created by millions of
  millions of Internet users), autonomously move
  and replicate.
• CEs make relationships with other CEs providing
  related services.
• Diverse behavior policies are created good
  behaviors survive, bad ones die, making system
  adaptable and evolvable.
• Let the Internet live its own life.

14
Outline

• Motivation and Biological Concepts in Bio-Net
• Emergent Behavior
• Evolution and Adaptation
• Peer to Peer Discovery in Bio Net
• Current Status of Other Subprojects
• Adaptation and evolution simulations
• Platform design
• Service emergence
• Conclusions

15
Peer to Peer Discovery in Bio Net

• Need for finding certain types of CEs
• information that soldiers collect in a combat
  situation
• information collected by fire fighters at the
  ground zero
• Under dynamic network changes
• CEs may move and die (soldiers move) (fire
  fighters move)
• Relationships may change
• Military applications
• Crisis management applications

16

• Community
• I managed NSF Net research program
• PIs also asked me about NSF Special Projects

Spec Proj
C
Net Res Prog
A
D
D
17

• Community
• I managed NSF Net research program
• PIs also asked me about NSF Special Projects

Spec Proj
C
Spec Proj
Net Res Prog
A
D
D
18

• Community
• Robust to dynamic network changes

Spec Proj
Spec Proj
Net Res Prog
A
C
D
D
19

• Community Creation

E
C
B
A
D
D
20

• Community Creation

E
C
Query for E
B
A
D
D
21

• Community Creation

E
C
Query hit
Query for E
B
A
D
D
22

• Community Creation

Add E
E
Reward
C
Query hit
Query for E
B
E
A
D
D
23

• Community Creation

Add E
Add E
E
Reward
E
C
Query hit
Query for E
B
E
A
D
D
24

• Community Creation

Add E
Add E
E
Reward
E
C
Query hit
Query for E
B
E
A
D
D
25

• Keyword Strength (usefulness)

E
Reward
E
C
B
E
A
D
D
26

• Keyword Strength (usefulness)

Increase strength for E
Increase strength for E
E
Increase strength for E
Reward
E
C
B
E
A
D
D
27

• Query forwarding

E
E
C
B
E
A
D
D
28

• Query forwarding

E
E
C
B
E
A
D
D
E
E
29

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
C
B
E
A
D
D
E
E
30

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
C
Query for E
B
E
A
D
D
E
E
31

• Query forwarding
• Probabilistic, proportional to keyword strength

Forward with higher probability
E
E
C
Query for E
B
E
A
D
D
E
E
32

• Query forwarding
• Probabilistic, proportional to keyword strength

E
E
C
Query for E
B
E
A
D
D
E
Forward with smaller probability
E
33

• Query forwarding
• Probabilistic, proportional to keyword strength
• Robust to dynamic network changes

E
E
C
Query for E
B
E
A
D
D
E
Forward with smaller probability
E
34

• Currently running simulations to investigate
• Scalability, Efficiency, Robustness to dynamic
  network changes

35
Outline

• Motivation and Biological Concepts in Bio-Net
• Emergent Behavior
• Evolution and Adaptation
• Peer to Peer Discovery in Bio Net
• Current Status of Other Subprojects
• Adaptation and evolution simulations
• Platform design
• Service emergence
• Conclusions

36
Evolution/Adaptation Simulations

• Behavior policy a weighted sum of factors

W2
W1
If
gt M
, then, migrate
W1, W2, M weights
Factors which affect behavior (migration)
direction a user request came from
cost of migration
etc.
37

• CE behavior mutation
• In weights
• In factors
• CE behavior crossover in factors

mutation
1001
1011
1001
1001
1001
1001
mutation
child factors
parent A
mutation
parent B
38
Simulation configuration
response time (mutation/crossover off)
response time (mutation/crossover on)
users movement

• Bio Net evolves
• Through mutation/crossover, CEs reduce response
  time to user requests.

39
Delay Graph of Aphid versus Static Servers
Monthly Cost of Aphid versus Static Servers
43.55
27.57
6 statically placed servers
Aphid
40
Bio-Networking Platform Designs
CE
CE
CEs communicate using FIPA ACL.
CE context references to bionet services.
CE Context
Bionet services general-purpose services (energy
management, relationship management, etc.)
Bionet Services
Bionet Container
Bionet container low-level operations for memory
and resource management.
Bionet Platform
Message Transport
Java VM
41

• Our platform design
• Being discussed at OMG as a possible standard for
  super-distributed systems
• Being implemented by NTT

42
Service Emergence

• A group of CEs collectively providing an
  application
• Developed a grouping algorithm based on adjusting
  relationship strength based on user feedback
• Empirically evaluating the algorithm through
  implementation of simple applications (with NTT)

43
Outline

• Motivation and Biological Concepts in Bio-Net
• Emergent Behavior
• Evolution and Adaptation
• Peer to Peer Discovery in Bio Net
• Current Status of Other Subprojects
• Adaptation and Evolution Simulations
• Platform design
• Service emergence
• Conclusions

44
Conclusions

• Bio Net is a new paradigm for scalable network
  applications
• Our accomplishments
• Simulator for bio net
• Algorithms
• Discovery algorithms
• Service emergence algorithms
• Diversity creation algorithms
• Platform design (and implementation)
• Application design (and implementation)
• Standard activities