WORKLOAD SCHEDULING IN MULTI-CORE PROCESSORS

• US 20090271141A1
• Filed: 04/25/2008
• Published: 10/29/2009
• Est. Priority Date: 04/25/2008
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Gross, Kenny C., Urmanov, Aleksey M., Whisnant, Keith A., Coskun, Ayse K.
• Granted Patent: US 7,716,006 B2
• Time in Patent Office: Days
• US Class Current: 702/132
• CPC Class Codes: G06F 1/206 comprising thermal management
• CPC Class Codes: G06F 9/5033 considering data affinity
• CPC Class Codes: G06F 9/5094 where the allocation takes into account power or heat criteria power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A computer system that schedules loads across a set of processor cores is described. During operation, the computer system receives thermal measurements from sensors associated with the set of processor cores, and removes noise from the thermal measurements. Then, the computer system analyzes thermal properties of the set of processor cores based on the thermal measurements. Next, the computer system receives a process to be executed, and schedules the process to be executed by at least one of the processor cores based on the analysis. This scheduling is performed in a manner that reduces spatial and temporal thermal variations in the integrated circuit.

Claims

1. A method for scheduling loads across a set of processor cores disposed on an integrated circuit, comprising:

receiving thermal measurements from sensors associated with the set of processor cores;

removing noise from the thermal measurements;

analyzing thermal properties of the set of processor cores based on the thermal measurements;

receiving a process to be executed; and

scheduling the process to be executed by at least one of the processor cores based on the analysis, wherein the scheduling reduces spatial and temporal thermal variations in the integrated circuit.

2. The method of claim 1, wherein the thermal properties include energy consumption.

3. The method of claim 1, wherein the thermal properties include temperature.

4. The method of claim 1, wherein the scheduling attempts to map related processes to processor cores that share the same memory.

5. The method of claim 1, wherein the scheduling is based on thermal properties of the set of processor cores while previous processes were executed.

6. The method of claim 1, wherein the scheduling facilitates load balancing across the set of processor cores.

7. The method of claim 1, wherein the thermal measurements include continuous telemetry measurements that are received from sensors disposed on the integrated circuit in proximity to the set of processor cores.

8. The method of claim 7, wherein removing noise from the thermal measurements involves validating the thermal measurements by comparing thermal measurements from the given sensor with predictions of a regression model that is based on measurements of other sensors that have a predetermined relationship with the thermal measurements from the given sensor.

9. The method of claim 8, wherein the regression model includes a multivariate state estimation technique (MSET).

10. The method of claim 8, wherein the predictions are used instead of the thermal measurements if the comparison indicates an error in the thermal measurements.

11. The method of claim 1, wherein a regression model is used to reduce noise in the thermal measurements without adding bias.

12. The method of claim 11, wherein the regression model includes Tikhonov regularization.

13. The method of claim 1, wherein the thermal properties are estimated based, at least in part, on parameters of a computer system that includes the integrated circuit;

and

wherein the parameters are determined using virtual sensors in an operating system of the computer system.

14. The method of claim 1, wherein analyzing the thermal properties involves using a thread which is separate from threads associated with other processes that are executed in the set of processor cores.

15. The method of claim 1, wherein the thermal measurements are received during a time interval;

and

wherein the analysis of the thermal properties is based on thermal measurements during the time interval.

16. The method of claim 1, wherein the scheduling is probabilistic;

and

wherein a probability of assigning processes to a given processor core during a time interval is determined based on one or more thermal properties of the given processor core during the time interval and a dynamic weight associated with the given processor core.

17. The method of claim 16, wherein the probability of assigning processes is determined using an adaptive-random technique.

18. The method of claim 16, wherein the one or more thermal properties includes temperature.

19. A computer-program product for use in conjunction with a computer system, the computer-program product comprising a computer-readable storage medium and a computer-program mechanism embedded therein for scheduling loads across a set of processor cores disposed on an integrated circuit, the computer-program mechanism including:

instructions for receiving thermal measurements from sensors associated with the set of processor cores;

instructions for removing noise from the thermal measurements;

instructions for analyzing thermal properties of the set of processor cores based on the thermal measurements;

instructions for receiving a process to be executed; and

instructions for scheduling the process to be executed by at least one of the processor cores based on the analysis, wherein the scheduling reduces spatial and temporal thermal variations in the integrated circuit.

20. A computer system, comprising:

a set of processor cores disposed on an integrated circuit;

memory configured to store thermal measurements associated with a set of processor cores; and

a scheduler, wherein the scheduler is configured to execute instructions, and wherein the instructions include;

instructions for receiving the thermal measurements from sensors associated with the set of processor cores;

instructions for removing noise from the thermal measurements;

instructions for analyzing thermal properties of the set of processor cores based on the thermal measurements;

instructions for receiving a process to be executed; and

instructions for scheduling the process to be executed by at least one of the processor cores based on the analysis, wherein the scheduling reduces spatial and temporal thermal variations in the integrated circuit.