The Bio-Networking Architecture: Adaptation of Network
Applications through Biological Evolution Jun Suzuki and Tatsuya
Suda {jsuzuki, suda}@ics.uci.edu
http://netresearch.ics.uci.edu/bionet/ Dept. of Information and
Computer Science University of California, Irvine
Slide 2
Goals of the Simulation Study To show that the Bio-Networking
Architecture adapts to diverse network conditions through
behavioral evolution of autonomous cyber-entities (CEs) To show
that evolutionary mechanisms (diversity generation and natural
selection) allow CEs to increase their fitness to diverse network
conditions.
Slide 3
Cyber-Entity (CE) Each CE behavior policy consists of factors
(F), weights (W), and a threshold. If > threshold, then
reproduce. Example reproduction factors: StoredEnergyFactor
contributes to the tendency for CEs to reproduce more often when
they have enough energy. RequestRateFactor contributes to the
tendency for CEs to reproduce more often when they receive a large
number of service requests. RequestChangeRateFactor contributes to
the tendency for CEs to reproduce more often when request rate is
increasing. Behavior Attributes Body GUID energy level age
non-exec. data executable code Cyber-entity migration replication
reproduction pheromone emission resource sensing energy exchange
social networking relationship relationship list Each CE stores and
expends energy in exchange for performing service. for using
resources.
Slide 4
Evolutionary Mechanisms Diversity generation A CE behavior may
be implemented by a number of algorithms/policies Manual diversity
generation by human designers Automatic diversity generation
through mutation and crossover during replication and reproduction
Natural selection keeps entities with beneficial features alive CEs
that adapt to environment well will contribute more to evolution.
Energy used as a natural selection mechanism abundance induces
replication and reproduction scarcity induces death
Slide 5
Automatic Diversity Generation Weight and threshold values in
each behavior policy change dynamically through mutation. Mutation
occurs during replication and reproduction. Behavior Policy
Parameter Set weight 1 weight 2 threshold Migration Policy Params
weight 1 weight 2 Weight 3 threshold Reproduction Policy
Params...... When reproducing, a CE selects a mate whose fitness to
the current network condition is high. Fitness is a function of
distance to users, response time to user requests, and energy
utility. A child CE inherits different behaviors from different
parents through crossover. Behavior Policy Parameter Set weight 1
weight 2 threshold Migration Policy Params weight 1 weight 2 Weight
3 threshold Reproduction Policy Params Behavior Policy Parameter
Set weight 1 weight 2 threshold Migration Policy Params weight 1
weight 2 Weight 3 threshold Reproduction Policy Params Behavior
Policy Parameter Set weight 1 weight 2 threshold Migration Policy
Params weight 1 weight 2 Weight 3 threshold Reproduction Policy
Params parents reproduced child
Slide 6
Example Simulation Results Investigates the impact of
mutation/crossover by comparing fitness of 2 populations of CEs;
one with mutation/crossover, and the other without
mutation/crossover Observation: Mutation/crossover allows CEs to
gradually shorten response time to user requests and reduce
distance to users. response time to user requests
(mutation/crossover on) response time to user requests
(mutation/crossover off) users movement Network configuration hop
counts between CEs and users (mutation/crossover on) hop counts
between CEs and users (mutation/crossover off)
Slide 7
Investigates fitness under different distributions of resource
cost. (Config. 1) Same resource cost on all the platforms (Config.
2) Different resource costs on different platforms Energy utility
resource cost Observation: CEs gradually shorten response time to
user requests in both config 1 and 2. The number of platforms
hosting CEs approaches toward 1 in config 1. This does not happen
in config 2. This means that CEs avoid to move to platforms whose
resource cost is high. CEs increase energy utility in config 2 than
in config 1. This means CEs save their energy in config 2 by
running on platforms whose resource cost is low. # of platforms
hosting CEs Response time