Mass space information-distribution type processing method based on free market model

Abstract

The method includes steps: resources i.e. executing hosts put forward applications for joining cluster to server; receiving task submitted from task presenter, server evaluates executive price for each task, carrying out dispatching assignment for executive task according to principle of minimum price; when task waited to execute is detected in own task storage space, idle executing host in distributed cluster carries out process of executing benefit for all candidate tasks, and carrying out sorting assignment for candidate tasks according to executing benefit; based on description of task, executing host requests program files and data files to pointed file server, and file server sends files needed by task to executing host; when task is completed, executing host sends output files to the task presenter. The invention realizes instant property of market information, reduces burden at server end so as to raise system performance.

Description

Technical field [0001] The invention relates to a method in the field of information processing technology, in particular to a method for distributed processing of massive spatial information based on a free market model. Background technique [0002] The problems of spatial information processing have the following characteristics: 1) Large amount of processing. Spatial information processing is generally large-scale image processing, and the processing volume is very large. 2) The amount of data is large. The data volume of spatial information is terabytes. 3) The processing form on each data point is the same (except for the boundary). If the area is divided into several small areas, the problem can be reduced to several smaller-scale sub-problems about small areas. 4) The interaction between variables is local. In the processing of each data point, only the value of the neighboring point within a small distance is used. Among the above features, features 1) and 2) contribute ...

AI Technical Summary

Problems solved by technology

Spatial information processing is generally large-scale image processing, and the processing capacity is very large.

2) Large amount of data

However, this method uses a prediction mechanism for resource status when assigning tasks. The disadvantages of the prediction method are obvious: the principle of most prediction methods is based on the theory of linear steady systems, and the status value of resources can only be obtained within a certain time range. It is approximately considered to be a stationary random process. As far as the overall situation is concerned, it is undoubtedly time-varying. Therefore, the prediction also has the problem of detection frequency. If the detection interval is too short, it will have a great impact on the system. Otherwise, it is difficult to guarantee Accuracy of Predicted Values

Moreover, all processing tasks are allocated on the server side. For processing massive spatial information, the burden on the server is very heavy. There must be an improved method to reduce the burden on the server and compensate for the error of the prediction mechanism.

Images