Current Projects - Information System Thermodynamics

Objective

Develop a model to predict the behavior of information systems

Approach

  • Represent an information system as two sets: devices and information
  • Represent the devices with a directed graph of processing and storage device vertices connected by communications device arcs
  • Label both vertices and arcs of the device graph with the device physical volume, device capacity, instantaneous device load, sets of the operations that the device can execute and the energies the device dissipates when it executes those operations; the devices of an information system determine the physical limits of the computational environment within which the information of the information system executes and therefore exists
  • Represent the information in an information system with a directed bipartite graph of a set of model properties connected to a set of model dependencies
  • Label the model properties with the instantaneous property value, property representation size and a set of all admissible property values
  • Label model dependencies with the dependency representation size and sets of independent variable properties, dependent variable properties, mappings from all admissible independent variable values to all possible dependent variable values, executable operations and computational work that must be performed to execute each operation
  • Map the models represented into the information system devices and initiate execution; at any moment in time, the information system state can be defined by the values of all of the properties of all of the instances of all of the models that the information system represents; similarly, at any moment in time, the next information system state is determined by the set of dependencies that the information system devices are currently executing; execution of this set of dependencies changes the information system state
  • Represent the set of all the possible property states that the information system dependencies could generate from all of the possible input property states as a directed graph with accessible property states as vertices and executable transitions between those states as arcs; each dependency represents the mapping between input property states (i.e., independent variable values) and output property states (i.e., dependent variable values); this mapping defines a set of transitions between property states
  • Derive approximate relationships between the properties of the device, model and state space graphs from thermodynamic considerations

Results

Derived several relationships between device capabilities and information representation characteristics

  • instantaneous device load ≤ device capability for all information system devices
  • instantaneous communications device load = sum of property sizes communicated to and from an executing dependency / dependency execution time
  • instantaneous storage device load = sum of property sizes for all model instances + sum of model dependency sizes
  • dependency execution time = dependency computational work / executing processing device capacity
  • instantaneous processing device load = sum of computational work for all executing dependencies
  • information system model complexity = size of its state space graph
  • information system volume = total size of the dependencies and property instances represented
  • instantaneous storage device load * instantaneous communications device load = constant * rate of property state change * model complexity

Work continues to explore the relationships between information system properties and new results will be reported when available.

Conclusions

This work

  • Identifies small set of observable macroscopic properties of all information systems;
  • Characterizes the dependencies between those properties; and
  • Establishes foundation for measuring the state and behavior of information systems and for characterizing the capabilities of specific types of information systems including intelligent systems.