New Classical Economic Based Resource Allocation Strategy And Method In Grid Computing Environment

Research on resource allocation is one of the major topics within computer theoretic field. Most traditional methods solve the problem with engineering mechanisms, which allocate resources via global computation. These mechanisms are effective in unified systems, but not applicable for grid systems. The resource allocation problem of distributed computing is NP-complete problem. The computation complexity will increate dramatically as the system scales up. Although some intelligent optimization algorithms are introduced, their actual effectiveness are not quite good due to the difficulty to find the optimal solution. Algorithms based on marketing economics show some promising characteristics, but hard provide real-time solution. Due to the specific difficulties of resource allocation under grid environment, none of current analysis and solutions is well established.In order to avoid the complexity of the analysis algorithm itself, this paper tries to solve resource allocation problem under grid environment by introducing new classical economics theory. Based on the principle that different node has respective comparative advantage on certain request, the idea of the new model is to distribute request onto the node which has comparative advantage to carry out the task. The new method uses both the division of work and specialization theory of classical economics and the inframarginal analysis of new classical economics. This method provides an analysis framework aiming to enhance system performance. It reduces the analysis complexity by combining macro and micro economics into a unified analysis framework. It is different from the Walrasian General Equilibrium based resource pricing model and the Nash Equilibrium based auction model. The presented method focuses not only on resource allocation itself, as the above mentioned two models do, but also on the resource allocation structure as well. It simplifies the complexity of the algorithm, reduces the solution space, effectively restricts solution explosion.The new classical economics based resource allocation method finds the node with maximum effectiveness and selects proper computing structure according exogenous comparative advantage theory. It turns the traditional difficulty of grid resource allocation problem into the problem of dispatching tasks to nodes with maximum effectiveness and dynamically network structure tuning. It means market competition can not only distribute resource effectively (corner equilibrium), but also select distribution structure (global equilibrium), therefore reach global Pareto Optimality. The method presented in this paper aims to enhance global performance and effectively allocate resources. It is a highly autonomous resource allocation method. Each node makes allocation strategy based on local information; therefore, the information explosion problem of traditional methods can be avoided. Different from intelligent optimization algorithms, this method is highly predictable as it makes allocation strategy via computation.This paper explains the inframarginal analysis based resource allocation method from three perspectives:the economic analysis model of grid resource, the resource allocation algorithm and the simulation test. The major works of this paper are:(1) Presents a grid resource economic model based on new classical economics theory, and provides economic analysis on the resource allocation within grid environment. By establishing the producer-consumer system within grid environment, and defining various consumption and transaction costs, candidate optimal corner solutions are found via multistep method under Walras’system and Kuhn-Tucker conditions, work division structure for different conditions are clarified, and comparative static analysis are performed.(2) Presents a work division decision strategy based on utility function and inframarginal analysis. The utility behavior is defined according Cobb-Douglas production function. Under condition of time endowment, common resource allocation model of grid environment are set up based on consumer choice and comparative advantage theory. Utility function, service vector and transaction vector of service nodes can therefore be defined. The procedure of model setup and solution acquisition is illustrated with the example of distributing a specific request onto two available nodes. Division of work and specialization can then further enhance global efficiency by distributing specific request onto proper node. Inframarginal analysis is applied on allocation details so that each step can be quantified.(3) Presents a method to select division structure and perform comparative static analysis based on the grid resource allocation model. With the corner analysis of service ability and exchange index of each division structure, three division structure of optimal resource allocation are obtained, and allocation algorithm is provided respectively. Each node is carrying over the task on which it has comparative advantage, resulting enhanced global system efficiency. Simulation test provides two set of nodes and utility function indexes. Analysis shows that while exchange efficiency increases, more services become exchangeable and more nodes are carrying over services they have comparative advantage on.(4) Presents a Pareto Optimality based allocation decision model. Conditions of selecting different structures are getting by performing decision analysis against different structures. Algorithms are set up to compute allocation structure. Simulation tests are performed on different allocation structure, and results are analyzed according characteristics of initial data and algorithm.The method presented in this paper is applicable for grid environment. It can response to unexpected changes and apply specific strategy according particular condition. It can also tune allocation strategy dynamically so that a high resource utilization rate can be maintained.