Resource Allocation in Physically Distributed System
using Non-Cooperative Game Theory

Abstract

HDCS(Heterogeneous Distributed Computing System) is a set of
interconnected computing nodes so as to oers sharing of computational power,
applications or network resources dynamically.Task scheduling is the crucial
part which guides the resource allocation in distributed computing systems.The
resource allocation in distributed computing system can be classied as centralized
and decentralized.
In this thesis decentralized dynamic resource allocation scheme in
distributed computing systems has been modeled as a non-cooperative game
among the computing nodes and the optimal solution of this non-cooperative
game is given by the nash equilibrium.The existing schemes assigns tasks to the
computing nodes which are having more processing capability but in dynamic
environment its not sucient to schedule tasks based on the processing powers.
Resource availability along with the processing powers of computational resources
is considered to be of great signicance with regard to performance issue. In the
proposed scheme resource allocation is based upon the processing powers of the
computing nodes, resource availability(waiting time), communication delay and
the execution times of tasks at dierent computing nodes. This scheme minimizes
makespan of the tasks while improving the fairness of the HDCS. The proposed
non-cooperative scheme is giving better makespans and fairness of this proposed
non-cooperative scheme is 90% i.e.,makespans at all the computing nodes in the
system is almost same.
Economical models have been widely used in the distributed system for
resource allocation,we have proposed an economical model proposed to minimize
the resource payments based on the current states of the resources instead of cost
based resource allocation. In the decentralized distributed system,we proposed an
agent based cost optimization with the objective for improving the revenue of the
consumers or users of the distributed system. Most importantly computational
price of the tasks is minimized by paying the price based on the resource availability
time(current state of the resource) instead of paying the standard prices. This
ivapproach provides economic incentives to resource owners by fair task scheduling.If
the system is unfair then the user can pay reduced price if the deadline of the tasks
are not critical. Both the resource owner agent and consumer agent bargained with
each other based on the current state of the resources with an objective of revising
the price of the resources, which is economical to both the resource owners and
consumers. All the proposed schemes has been veried through simulation against
the existing game theory approaches.