Managing processor usage of a physical host configured for hosting computing instances

Patent metadata

Field	Value
Publication number	US-9959188-B1
Application number	US-201514791120-A
Country	US
Kind code	B1
Filing date	Jul 2, 2015
Priority date	Jul 2, 2015
Publication date	May 1, 2018
Grant date	May 1, 2018

Abstract

Technology for detecting processor usage of a physical host hosting computing instances is provided. In one example, a method may include identifying a location of a hardware register of a physical host containing data in a model specific register. The physical host hosts a plurality of computing instances using a hypervisor. The method may include determining power consumption by a processor of the physical host for the computing instances from the data. An upper level of acceptable processor power consumption may be identified. Processor usage may be determined based on the power consumption determined and the upper level of acceptable processor power consumption. The processor usage may be provided as data through the hypervisor.

First claim

The invention claimed is:

1. A physical host with a computing instance that is configured to detect processor usage of the physical host hosting a plurality of computing instances, comprising:
a processor of the physical host;
a memory in electronic communication with the processor;
instructions stored in the memory, the instructions being executable by the processor to:
read data from a model specific register located in a hardware register memory of the physical host;
determine, using a computing instance executing on the physical host, power consumption by the processor of the physical host from the data, wherein the power consumption is caused by a plurality of computing instances, and the data is received at the computing instance through a hypervisor;
identify, using the computing instance, a thermal design point of the processor, the thermal design point being indicative of an upper level of acceptable processor power consumption;
determine, using the computing instance, processor usage based on the power consumption as compared to the thermal design point; and
implement a noisy neighbor remediation policy based on the processor usage.

2. The computing instance of claim 1 , wherein the computing instance is configured to implement the noisy neighbor remediation policy by migrating the computing instance to a different physical host when the processor usage is above a predetermined threshold.

3. The computing instance of claim 1 , wherein the computing instance is configured to implement the noisy neighbor remediation policy by initiating a turbo boost of the processor when the processor usage is above a predetermined threshold.

4. The computing instance of claim 1 , wherein the computing instance is configured to implement the noisy neighbor remediation policy by canceling initiation of software processes on the computing instance when the processor usage is above a predetermined threshold.

5. The computing instance of claim 1 , wherein the data from the model specific register is stored as metadata in the hypervisor for the computing instance.

6. The computing instance of claim 1 , wherein the computing instance is configured to make a modification to a scheduling policy for placing a new computing instance based on the processor usage, wherein the modification is the noisy neighbor remediation policy.

7. A computer-implemented method, comprising:
identifying a location of a hardware register of a physical host containing data in a model specific register defined to track power consumption, using a processor, the physical host hosting a plurality of computing instances;
determining, using a computing instance executing on the physical host, power consumption by a processor of the physical host for the plurality of computing instances from the data, wherein the data is received at the computing instance through a hypervisor,
identifying, using the computing instance, an upper level of acceptable processor power consumption;
determining, using the computing instance, processor usage based on the power consumption determined and the upper level of acceptable processor power consumption, using the processor;
performing a processor usage action in response to the processor usage.

8. The method of claim 7 , further comprising implementing a noisy neighbor remediation policy based on the processor usage as the processor usage action.

9. The method of claim 7 , further comprising identifying the upper level of acceptable processor power consumption by identifying a thermal design point of the processor.

10. The method of claim 9 , further comprising identifying the upper level of acceptable processor power consumption by identifying an acceptable power consumption increase over the thermal design point.

11. The method of claim 7 , further comprising migrating the computing instance to a different physical host when the processor usage is above a predetermined threshold as the processor usage action.

12. The method of claim 7 , further comprising initiating a turbo boost of the processor when the processor usage is above a predetermined threshold as the processor usage action.

13. The method of claim 7 , wherein reading the power consumption by the processor comprises reading the power consumption for a plurality of cores of the processor.

14. The method of claim 7 , further comprising monitoring the power consumption of the processor for the physical host over a predetermined period of time and determining whether to implement the processor usage action based on the processor usage.

15. The method of claim 7 , further comprising accessing the data in the model specific register of the physical host by one of the plurality of computing instances by accessing a Uniform Resource Locator exposing metadata of the physical host including the data in the hardware register.

16. The method of claim 7 , further comprising determining an identity of the processor, and identifying the upper level of acceptable processor power consumption based on the identity of the processor.

17. A non-transitory computer-readable medium comprising computer-executable instructions which implement a system, comprising:
a memory accesser configured to provide access to register data memory which is configured to store register data of a model specific register of a processor of a physical host, wherein the memory accesser is executed on a hypervisor for the physical host;
a processor identifier, within a computing instance executing on the physical host, configured to identify a processor of the physical host and identify information about the processor, the information including at least a thermal design point of the processor;
a power usage determiner, within the computing instance, configured to determine a processor usage based on processor power consumption data included in the register data and based on the thermal design point of the processor, wherein power consumption is caused by a plurality of computing instances, and the processor power consumption data is received at the computing instance through the hypervisor; and
a policy engine configured to implement a noisy neighbor remediation policy when the processor usage is above a predetermined threshold.

18. The computer-readable medium of claim 17 , further comprising a migrater configured to migrate a computing instance hosted on the physical host to a second physical host according to the noisy neighbor remediation policy when the processor usage is above the predetermined threshold.

19. The computer-readable medium of claim 17 , further comprising a monitor configured to monitor the processor usage over time and determine an average processor usage, wherein the policy engine implements the noisy neighbor remediation policy when the average processor usage is above a predetermined threshold.

20. The computer-readable medium of claim 17 , wherein the power usage determiner is configured to calculate availability of the processor by estimating a processor usage as a percentage of total available processor usage derived from the thermal design point and the processor power consumption data.

Assignees

• Amazon Tech Inc

Inventors

• Krishnan Karthikeyan

Classifications

• G06F2009/45579

  I/O management, e.g. providing access to device drivers or storage · CPC title

• G06F9/5077

  Logical partitioning of resources; Management or configuration of virtualized resources (specific details on emulation or internal functioning of virtual machines G06F9/455) · CPC title

• G06F11/3062 Primary

  where the monitored property is the power consumption (power management in a computing system G06F1/3203) · CPC title

• G06F9/5094

  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 title

• G06F11/3024

  where the computing system component is a central processing unit [CPU] · CPC title