Resource Allocation Based On New Classic Economics

As the development and widely usage of distributed systems, issues of resource allocation are becoming increasingly remarkable. Traditional resource allocating methods are largely implemented through control technique applied in areas of engineering, which is allocating the resource into appropriate nodes. This method is efficient in a sole system, for that the mount of resources in a sole system are small enough to synchronizes with each other. However it is proved to be a NP problem in a distributed system, which means the size of computing will become significantly large as the scale of system increase. In fact, it is inaccessible if it is simply brought into a distributed system, thus a serial of improved measures have been researched and implemented in real systems. Among these policies, intelligent optimization and market-based resource allocating methods are comparatively outstanding. However, it is difficult for intelligent optimizations to ensure the solutions are as optimal as expected, which is even so if the system is unstable. The market-based methods are proposed to introduce microeconomics into distributed systems, which is proved to be quite successful for that the market is originally a distributed organization. Unfortunately, its executing periods, according to recent researches, is too long to fit some time-restrict systems. These measures are still in researching stage, and largely based on neoclassic economics.This thesis, with supporting of theory about new classical economics, is mainly arguing the problem of resource allocation under the environment of distributed systems. This method is largely concerning about how to improve the performance of the whole system, simultaneously maximizing the utility of each client. In these processes, effective policies of resource allocation are absolutely important to achieve this goal. The steps are:(1) The combination of topological structure about the distributed system and abilities of nodes handling requests, associating with some additional computing, determine which allocating structure it is laying on. The differences of net bandwidth between nodes imply that there are different transaction efficiencies among those nodes, which means varietals transaction coefficients existed. Each of nodes has different capacities in providing services for clients, which present comparative advantage is inherent for services between two nodes. These two external factors decide the structure of allocation between nodes about two services.(2) Under the definite structure in the previous step, optimal points can be obtained through derivation, thus the amount of resource allocation between these two nodes is aware. This step is accomplished under helps of marginal analysis in neoclassic economics.(3) Allocate the requests into these two nodes according to the second step. And allocate all the requests into these two nodes similarly.After these steps, each class of request is now allocating to the nodes which have comparative advantage in dealing with it, and each node is maximizingly providing services it have comparative advantage in handling. Furthermore, requests from each service net are contained in the specialization of work. Totally, this idea is translating the problems of resource allocation into issues of specialization of work, and then applies inframarginal analysis to get the optimal solutions.