Dynamic data and compute resource elasticity

Patent metadata

Field	Value
Publication number	US-9600255-B2
Application number	US-90069610-A
Country	US
Kind code	B2
Filing date	Oct 8, 2010
Priority date	Oct 8, 2010
Publication date	Mar 21, 2017
Grant date	Mar 21, 2017

Abstract

An execution environment in a computer system provides dynamic data and compute resources elasticity for user code to improve execution efficiency. The execution environment translates the user code into a runtime agnostic representation with a set of tasks. For each task, the execution environment determines a level of concurrency for executing the task based on the size of the set of input data for the task, the amount of compute resources available at the time of invocation of the task, and any context-sensitive heuristics provided by the user code.

First claim

What is claimed is:

1. A computer readable storage medium storing computer-executable instructions that, when executed by a computer system, cause the computer system to perform a method comprising:
dynamically binding user code to a selected execution runtime, selected from a set of at least two different execution runtimes;
delineating a programming model for the user code from the set of at least two different execution runtimes thereby the same user code may be used to target the at least two different execution runtimes without incurring an overhead of including runtime-specific code for each execution runtime to provide automatic scaling of execution of the user code from a single machine to multiple machines;
determining a first level of concurrency for a first task from the user code based on a first size of a first set of input data for the first task and a first amount of compute resources of the computer system available for executing the first task; and
invoking a number of instances of the first task in dependence on the first level of concurrency.

2. The computer readable storage medium of claim 1 , the method further comprising:
determining the first level of concurrency for the first task on invocation of the task.

3. The computer readable storage medium of claim 1 , the method further comprising:
determining the first level of concurrency in response to the first set of input data corresponding to the first task becoming available.

4. The computer readable storage medium of claim 1 , the method further comprising:
dispatching the instances of the first task to the compute resources.

5. The computer readable storage medium of claim 1 , the method further comprising:
determining the first level of concurrency for the first task based on at least one context-sensitive heuristic provided by the user code.

6. The computer readable storage medium of claim 1 , the method further comprising:
determining a second level of concurrency for a second task from the user code based on a second size of a second set of input data for the second task and a second amount of compute resources of the computer system available for executing the second task; and
invoking a number of instances of the second task in dependence on the second level of concurrency.

7. The computer readable storage medium of claim 1 , the method further comprising:
translating the user code into a runtime agnostic representation that includes the first task.

8. The computer readable storage medium of claim 1 , wherein the user code inherently expresses a set of tasks that include the first task and a set of interactions of the set of tasks.

9. The computer readable storage medium of claim 1 , wherein the computer system includes one of a single machine, a cluster of machines, or a cloud of machines.

10. A computer readable storage medium storing computer-executable instructions that, when executed by a computer system, cause the computer system to perform a method comprising:
dynamically binding user code to a selected execution runtime, selected from a set of at least two different execution runtimes;
delineating a programming model for the user code from the set of at least two different execution runtimes thereby the same user code may be used to target the at least two different execution runtimes without incurring an overhead of including runtime-specific code for each execution runtime to provide automatic scaling of execution of the user code from a single machine to multiple machines;
dispatching a first number of instances of the first task from the user code in dependence on a first level of concurrency determined from a first size of a first set of input data for the first task and a first amount of compute resources of the computer system available for executing the first task; and
dispatching a second number of instances of the second task from user code in dependence on a second level of concurrency determined from a second size of a second set of input data for the second task and a second amount of compute resources of the computer system available for executing the second task.

11. The computer readable storage medium of claim 10 , the method further comprising:
determining the first level of concurrency in response to the first set of input data corresponding to the first task becoming available; and
determining the second level of concurrency in response to the second set of input data corresponding to the second task becoming available.

12. The computer readable storage medium of claim 10 , wherein the first level of concurrency differs from the second level of concurrency.

13. The computer readable storage medium of claim 10 , the method further comprising:
translating the user code into a runtime agnostic representation that includes the first and the second tasks.

14. The computer readable storage medium of claim 10 , wherein the user code inherently expresses a set of tasks that include the first and the second tasks and a set of interactions of the set of tasks.

15. The computer readable storage medium of claim 10 , the method further comprising:
determining the first level of concurrency for the first task based on at least a first context-sensitive heuristic provided by the user code; and,
determining the second level of concurrency for the second task based on at least a second context-sensitive heuristic provided by the user code.

16. The computer readable storage medium of claim 10 , wherein the computer system includes one of a single machine, a cluster of machines, or a cloud of machines.

17. A method performed by at least one computer system, the method comprising:
dynamically binding user code to a selected execution runtime, selected from a set of at least two different execution runtimes;
delineating a programming model for the user code from the set of at least two different execution runtimes thereby the same user code may be used to target the at least two different execution runtimes without incurring an overhead of including runtime-specific code for each execution runtime to provide automatic scaling of execution of the user code from a single machine to multiple machines;
translating the user code into a runtime agnostic representation that includes a first task, the user code inherently expressing a set of tasks that include the first task and a set of interactions of the set of tasks;
in response to a first set of input data corresponding to the first task becoming available, determining a first level of concurrency for the first task based on a first size of the first set of input data for the first task, a first amount of compute resources of the computer system available for executing the first task, and at least one context-sensitive heuristic provided by the user code; and
invoking a number of instances of the first task in dependence on the first level of concurrency.

18. The method of claim 17 , further comprising:
dispatching the instances of the first task to the compute resources.

19. The method of claim 17 , further comprising:
determining a second level of concurrency for a second task from the user code based on a second size of a second set of input data for the second task and a second amount of compute resources of the computer system available for executing the second task; and
invoking a number of instances of the second task in dependence on the second level of concurrency.

20. The method of claim 17 , wherein the computer system includes a s

Assignees

• Varadarajan Krishnan
• Chu Michael L
• Microsoft Technology Licensing Llc

Inventors

• Varadarajan Krishnan
• Chu Michael L

Classifications

• G06F9/5033

  considering data affinity · CPC title

• G06F8/443 Primary

  Optimisation · CPC title

• G06F8/456 Primary

  Parallelism detection · CPC title