Font Size: a A A

Researches On Grid Resource Allocation Strategy Based On Incomplete Information Game

Posted on:2010-04-08Degree:MasterType:Thesis
Country:ChinaCandidate:L XuFull Text:PDF
GTID:2178360302960611Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
As a kind of large area oriented distributed computing based on the high speed network, the grid computing integrits different calculating and memory resource which locate geographily and distributedly to complete the highly complex computing problems one single PC or a group of PCs can not do before.The field of resource management is always a tough task in grid computing.This paper refers to the theory of micro-economic and statistic,and bring a novel method that could optimize the grid resource assignment,the target is to improve the efficiency of allocation and validate the correction of this method,there are three aspect in this paper,including the resource allocation model,theory provation,experiment simulation,the following is the exact work this paper contains:analyze the OGSA and five layer hierarchical configuration,and compare three kinds of grid resource managment model with emphasising the marketing model's fair,high efficient, convenient charactors,then the clear description of several models used to solve the problem of grid resource allocation in the marketing grid environment will be given,at last,this paper summeriza the main problem and the limitaion during the grid resource assigning process.after researching those marketing models,one novel model named "Multi-Winners " grid resource auction model.It means numeral users compete for one kind of resource in grid.In case the resource provider can not decide the clear price of the resource and the users are independent to each other,the solvation this paper offers could make the final price of the resource fairly and neatly,which is the progress of solving the nash equilibrim.Its fastness, fairness and reliablity quite fit to the grid environment.Apply hidden markov prediction model into grid customer's price prediction firstly,set up descries time HMM,and predict the price per piece of resource,on condition that every customer's requirement of resource t-1 pieces of time before {r1,r2,...,rt-1} could be known,in order to compute the probabality for every user to get the resource they need.HMM theory give the proper description to grid with charactors of dynamic and randomness,in result that the efficiency of resource allocation is improved.The proving progress will be given to show the feasiblity of MWAM and the capability to maximize the system profit.As the economica based theory,MWAM is proved correctly from four aspects containing the exsistence of nash equilibrium,inspire-tollerating,personal reasonablity,system profit. Through the analysis of provation in math and the result of experiment,the MWAM this paper refers to satisfy the theorys in economic,and not only optimize the profit of the whole system and both the users and resource provider,improve the resource efficiency,but also show its useful and utilizable character to the grid.
Keywords/Search Tags:Hidden markov prediction, Uncompleted information game, Resource allocation, "Multi-winners"auction
PDF Full Text Request
Related items