Method and apparatus to manage multi-computer demand
View Patent ↗A method, multi-computer media and apparatus that uses an economics model to manage the demand and the resource satisfying that demand for multi-computer memory. The present invention quantifies demand as a function of space, and computer resource as a function of time, so that a computational system can meet an application demand.
1. A method to determine the memory requirements of an application running in parallel on the system, comprising the steps of:
inputting a model and initializing a computational domain;
calculating a data density for each control element;
calculating demand cost for each sub-domain;
minimizing the difference in average demand cost;
ranking the processors by value; and
generating a data ownership table and frame file.
2. The method of claim 1 , wherein the model is a discretized system model of a physical system.
3. The method of claim 1 , wherein initializing a computational domain also comprises dividing the domain into a number of equal sized geographic sub-domains with respect to the space coordinates of the model.
4. The method of claim 3 , wherein initializing a computational domain also comprises dividing the sub-domains into an integer fraction of rows and/or columns.
5. The method of claim 1 wherein data density within the sub-domains is represented by linear, area, and volume data demand functions.
6. The method of claim 1 wherein the demand cost is an area under a data density curve and is calculated by a numerical integration method.
7. The method of claim 1 wherein minimizing the difference in average demand cost also comprises adjusting the sub-domain size by “moving” bisectors.
8. The method of claim 7 wherein minimizing the difference in average demand cost also comprises recomputing data density based on a grid in each size adjusted sub-domain.
9. A system of networked computers having a plurality of processors and an operating system for executing a target parallel application process using at least a subset of said plurality of processors, wherein said system includes a method to determine the memory requirements of an application running in parallel on the system, said method comprising:
inputting a model and initializing a computational domain;
calculating a data density for each control element;
calculating demand cost for each sub-domain;
minimizing the difference in average demand cost;
ranking the processors by value; and
generating a data ownership table and frame file.
10. The method of claim 9 , wherein the model is a discretized system model of a physical system.
11. The method of claim 9 , wherein initializing a computational domain also comprises dividing the domain into a number of equal sized geographic sub-domains with respect to the space coordinates of the model.
12. The method of claim 11 , wherein initializing a computational domain also comprises dividing the sub-domains into an integer fraction of rows and/or columns.
13. The method of claim 9 wherein data density within the sub-domains is represented by linear, area, and volume data demand functions.
14. The method of claim 9 wherein the demand cost is an area under a data density curve and is calculated by a numerical integration method.
15. The method of claim 9 wherein minimizing the difference in average demand cost also comprises adjusting the sub-domain size by “moving” bisectors.
16. The method of claim 15 wherein minimizing the difference in average demand cost also comprises recomputing data density based on a grid in each size adjusted sub-domain.
17. A signal-bearing medium tangibly embodying a program of machine-readable instructions executable by a digital processing apparatus to determine the memory requirements of an application running in parallel on the system, said machine-readable instructions comprising:
inputting a model and initializing a computational domain;
calculating a data density for each control element;
calculating demand cost for each sub-domain;
minimizing the difference in average demand cost;
ranking the processors by value; and
generating a data ownership table and frame file.
18. The method of claim 17 , wherein the model is a discretized system model of a physical system.
19. The method of claim 17 , wherein initializing a computational domain also comprises dividing the domain into a number of equal sized geographic sub-domains with respect to the space coordinates of the model.
20. The method of claim 19 , wherein initializing a computational domain also comprises dividing the sub-domains into an integer fraction of rows and/or columns.
21. The method of claim 17 wherein data density within the sub-domains is represented by linear, area, and volume data demand functions.
22. The method of claim 17 wherein the demand cost is an area under a data density curve and is calculated by a numerical integration method.
23. The method of claim 17 wherein minimizing the difference in average demand cost also comprises adjusting the sub-domain size by “moving” bisectors.
24. The method of claim 17 wherein minimizing the difference in average demand cost also comprises recomputing data density based on a grid in each size adjusted sub-domain.