Overview

The Bio-Networking Architecture is a biologically inspired network architecture where some of the biological principles are applied to develop scalable and adaptive network applications. One of the key biological concepts applied to the Bio-Networking Architecture is adaptation and evolution by natural selection. Analogous to biological entities in the real world biological systems, cyber-entities in the Bio-Networking Architecture follw simple behavior rules (e.g., replication, reproduction and migration). Cyber-entities invoke these behaviors based on local information (e.g., resource availability around them) to improve a service performance (response time, resource usages, etc).

Cyber-entities evolve behaviors over generations. When a cyber-entity replicates or reproduces with another cyber-entity, diverse behavior policies are created through mutation and crossover in behavior policies. On the other hand, there is a common resource called energy introduced in the Bio-Networking Architecture; when cyber-entities provide a service to users, they receive energy; cyber-entities also consume energy in order to survive (e.g., to invoke a behavior). Since cyber-entities who run out of energy are going to die, only energy-efficient cyber-entities (that satisfy users and also use energy resources efficiently) are retained in the system. Thus, network applications become adapted to network environments.

In this project, we particulary investigate the adaptation and evolution aspects of the Bio-Networking Architecture. We have developed a simulator and, through simulations, demonstrated that the designed evolution mechanisms further improve the emergent properties of the Bio-Networking Architecture such as scalability, adaptability and survivability.

For more information, please see a project summary. Also, several papers are available at our publications page.

Download simulator

Bio-Networking Evolution Simulator has been developed to investigate the adaptation and evolution aspects of the Bio-Networking Architecture. A cyber-entity in this simulator consists of a set of behavioral factors that determines when to reproduce, replicate, migrate, and die. Cyber-entities interact with a platform, other cyber-entities and users in their surrounding environment, and invoke the behaviors based on information locally available to them. When a cyber-entity replicates or two cyber-entities reproduce, the evolutionary mechanism applied in this simulator modifies a behavioral factor in new offspring, creating diverse behavioral cyber-entities. By natural selection mechanism and through successive generations, only beneficial cyber-entities are retained on the network, enabling a network application to well-adapt in changing environments. The Bio-Networking Evolution Simulator is designed to demonstrate such evolvability and adaptability in the Bio-Networking Architecture.

Screenshots: 1, 2, 3

• version 4.4, released 12/18/2002

  Major improvements since the last version:

  • Command line mode supported
  Numerical results obtained from a set of simulations are available from here (PDF)

• version 4.3, released 08/12/2002

  Major improvements since the last version:

  • Support simulations in a variety of network environments including user random walk, topological changes (probabilistic and deterministic), resource cost changes over time
  • Renew configuration files and support automatic generation of them
  • Refine factor values

• version 4.2, released 01/11/2001

• version 4.1, released 6/05/2001:

  You can download a design document from here(Gzip'd, 1,594KB. Released 4/05/2001). Major improvements since the last version:

  • Add bonus option to give ce more energy when ce does good job.
  • Add several output attributes to analyze which factor affect on CE's decision.

• version 4.0, released 4/05/2001:

  You can download a design document from here. Major improvements since the last version:

  • New input and output files
  • Improved evolution mechanism in cyber-entities
  • New factors introduced to cyber-entity's behaviors
  • Improved GUI configuration
  This simulator requires JDK1.2 or higher.

• version 2.2, released 3/7/2000.

  Includes simulation source code and execution environment (Gzip'd, 450KB)
  This simulator was used for the technical report TR00-03.

  Major improvements since the last version:

  • Improved cyber-entity relationship code so that the super-entity is less likely to partition.
  • Changed all of the migration, reproduction, and death factors to make their decisions based on percentage of maximum rather than a hard coded value.
  The simulator requires JDK1.2, which you can download from java.sun.com. The simulator also requires IBM's Graph Foundation Class, which you can download here.

• Previous versions of the simulator.

Contact

Tadashi Nakano (tnakano@ics.uci.edu)