Monash University
4664455_monash_110797.pdf (7.58 MB)

An optimization framework using multiple economic models in grid computing: a switching mechanism

Download (7.58 MB)
posted on 2017-02-17, 03:59 authored by Haque, S. M. Aminul
Grid computing solves computationally complex problems such as climate modelling in a cost effective and standardized way. It requires seamless collaboration of computational resources distributed across different administrative domains worldwide. However, distributed ownerships and heterogeneous (independent) nature of these resources impose a challenge to this collaboration. Since economic-based resource management approaches have been found efficient and sustainable for various distributed computing platforms such as Grid, significant efforts are being made to evaluate the effectiveness of various economic models for distributed resource management. Several economic models have been proposed for Grid computing based on both micro and macro-economic principles. In spite of the potential of economic-based resource management, there is no consensus on choosing a particular economic model for the Grid as different researchers have proposed different models in different times. Therefore, a comprehensive understanding about various economic models in the context of Grid computing is essential to discover the problem of choosing a common model. The primary contribution of this thesis in identifying this problem is the process of a survey on existing economic models in the Grid. The survey identifies that one model is different from another in terms of pricing methodology and working principle. Moreover, the survey claims the suitability of different models for different scenarios. For example, Bargaining economic model supports utility-based negotiation between a resource user and a resource provider (microeconomic), whereas Commodity Market Model is suitable for maintaining equilibrium between supply and demand of resources in the environment (macroeconomic). The major reason to this problem is the limitation of a single model to cope with large-scale dynamic characteristic of the Grid. This limitation demonstrates a need to analyze the effectiveness of different economic models in Grid resource management. Therefore, this thesis conducts an extensive experimental analysis on the five most widely proposed economic models in the Grid – Commodity Market, Bargaining, English Auction, Continuous Double Auction and Contract Net Protocol. The experimental results demonstrate the compatibility with existing literature that a single economic model is not suitable for all circumstances in a Grid’s life-cycle. A quantitative analysis on the performances of different economic models helps identify the regions (domains) where one model outperforms all the other models in different scenarios. This variation in performances shows the opportunity of designing an optimization framework through utilization of the potentials of different models in different scenarios based on the domains of strengths of the models. To facilitate the optimization process, an adaptive switching mechanism that dynamically switches from one economic model to another depending on the function needed to be optimized, has been developed. The roles and responsibilities of the Grid entities to adapt with changing scenarios (one model to another model) in a dynamic environment have been justified and presented. The thesis further provides formal definitions to these domains of strengths of individual models to ensure that the switching decision can be carried out without much delay and computational power. The effectiveness of the switching framework in distributed resource management has been evaluated through a series of experiments. The results of these experiments show that the switching model can bring promising outcomes in collaborating distributed resources in an economic Grid.


Campus location


Principal supervisor

Saadat Alhashmi

Year of Award


Department, School or Centre

Information Technology (Monash University Malaysia)


Doctor of Philosophy

Degree Type



Faculty of Information Technology

Usage metrics

    Faculty of Information Technology Theses


    No categories selected


    Ref. manager