There are no similar models in the Empirical Modelling
archive and therefore this project involves the
construction of an entirely new model. This study
will mainly focus on modelling supply and demand
dependencies within the virtual Grid economy of
computational resources. This revolves around the
fact that as demand for a particular resource
increases, this resource becomes more valuable and
consequently its access price should rise to
reflect this and to obtain the maximum possible
return for the resource owner. This price increase
also has the effect of balancing the supply and
demand, as the price rises there will be fewer
consumers willing to purchase the resource and so
an equilibrium state will be reached. Similarly if
a resource is underutilised the access cost of the
particular resource should be lowered to stimulate
demand. It is hoped that explicitly modelling these
supply and demand dependencies will help to
determine how the pricing of resources depends on
the current state of the resource trading occurring
in the Grid. This project will also examine the
effects of using different economic models as a
basis for the resource economy and determine which
pricing policies and models are most appropriate to
different situations. These economic models include
Commodity Market, Posted Price, Bargaining,
Auction, Monopoly, Oligopoly and Contract-Net based
approaches. The observables which the project will
be most concerned with are therefore how the access
cost of various resources varies throughout the
simulations but also where particular jobs are
executed in the overall Grid environment as this
will help to determine which setups achieve the
best load balancing and overall throughput in
different situations. To achieve these ambitious
aims it will therefore be necessary to model the
operation of different processors within the Grid
environment and how the availability of these
varies over time due to possible external load
factors, as well as the effects of different
overall resource configurations. It may also be
necessary to model the communication links between
different agents in order to produce a model which
more accurately reflects real world situations, as
Grid systems are likely to be deployed across
numerous different physical networks consisting of
a variety of different bandwidths and latencies. It
is hoped that the project will be able to make use
of notations such as LSD in the modelling of the
agents.
