METHOD FOR DYNAMIC LOAD BALANCING ON PARTITIONED SYSTEMS

• US 20100125845A1
• Filed: 01/26/2010
• Published: 05/20/2010
• Est. Priority Date: 12/29/2006
• Status: Active Grant

• Current Assignee: Kiran Panesar, Suresh Sugumar
• Original Assignee: Kiran Panesar, Suresh Sugumar
• Inventors: Panesar, Kiran, Sugumar, Suresh
• Granted Patent: US 7,849,286 B2
• Time in Patent Office: Days
• US Class Current: 718/1
• CPC Class Codes: G06F 2009/4557 Distribution of virtual machine instances; Migration and load balancing
• CPC Class Codes: G06F 2209/5022 Workload threshold
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/5088 involving task migration

Abstract

Methods, systems and apparatuses to dynamically balance execution loads on a partitioned system among processor cores or among partitions.

Claims

1. A computing apparatus, comprising:

a plurality of partitions, wherein each partition is allocated a subset of processor cores of the computing apparatus without overlap or double assignment of the same processor cores to more than one of the partitions and wherein each partition includes one or more virtual machines to execute via the subset of processor cores assigned to the partition;

a first computer instruction to determine a load status of the subset of processor cores allocated to a first partition and compare the load status against an upper threshold for the first partition;

a second computer instruction to identify a virtual machine among the one or more virtual machines executing on the subset of processor cores allocated to the first partition that is running a task which is not present on a native task list for the first partition; and

a third computer instruction to migrate the virtual machine running the task which is not present on the native task list for the first partition to a second partition when the load status of the first partition exceeds the upper threshold for the first partition.

2. The computing apparatus of claim 1, wherein the computing apparatus further comprises:

a fourth instruction to select the second partition from among a plurality of available partitions based on a comparison of load statuses for each of the respective plurality of available partitions.

3. The computing apparatus of claim 1, wherein the computing apparatus further comprises:

a fourth instruction to select the second partition from among a plurality of available partitions based on a comparison of VM processor core temperatures for each of the respective plurality of available partitions.

4. The computing apparatus of claim 1, wherein the first partition comprises a system partition to perform system tasks initiated on behalf of an end-user and not initiated directly by the end-user, and wherein the second partition comprises a user partition to perform general tasks initiated directly by the end-user.

5. The computing apparatus of claim 1, wherein the computer instructions are encoded in hardware, or software, or firmware, or processor core micro-code, or a combination thereof.

6. The computing apparatus of claim 1, further comprising:

a virtual machine monitor to generate the computer instructions for execution on the computing apparatus.

7. The computing apparatus of claim 6, wherein the first virtual machine monitor to manage the one or more virtual machines to execute on the first partition, and further comprising:

a second virtual machine monitor to manage a second one or more virtual machines to execute on the second partition.

8. The computing apparatus of claim 7, further comprising:

a communication channel to connect the first virtual machine monitor with the second virtual machine monitor to exchange data corresponding with the first and second partitions; and

wherein the communication channel comprises either a shared memory component, or a direct memory access component, or both.

9. A method comprising:

hosting a plurality of partitions on a single computing apparatus, wherein each of the plurality of partitions is allocated a subset of processor cores within the computing apparatus without overlap or double assignment of any of the processor cores to more than one of the plurality of partitions;

executing one or more virtual machines within each of the plurality of partitions of the computing apparatus, wherein each virtual machine executes via the subset of processor cores assigned to the respective partition in which the one or more virtual machines operate;

determining a load status of the subset of processor cores allocated to a first partition among the plurality of partitions within the computing apparatus and comparing the load status of the first partition against an upper threshold for the first partition;

identifying a virtual machine among the one or more virtual machines executing on the subset of processor cores allocated to the first partition which is running a task not present on a native task list for the first partition;

migrating the virtual machine running the task which is not present on the native task list for the first partition to a different partition within the computing apparatus when the load status of the first partition exceeds the upper threshold for the first partition; and

repeating the determining, identifying, and migrating until the load status of the first partition is below the upper threshold for the first partition.

10. The method of claim 9:

wherein the first partition comprises a system partition to perform system tasks initiated on behalf of an end-user, rather than initiated directly by an end-user of the computing apparatus; and

wherein migrating the virtual machine to a different partition within the computing apparatus comprises migrating the virtual machine to a user partition that performs general computing tasks initiated directly by the end-user.

11. The method of computing of claim 9, further comprising:

modifying the frequency at which the load status of the first partition is determined based on a current load status of the first partition.

12. The method of claim 9, wherein migrating the virtual machine comprises the virtual machine:

executing in a first state of pre-migration, wherein the virtual machine is executing as normal;

executing in a second state of reservation, wherein the virtual machine is executing on the first partition and corresponds to a newly initialized virtual machine on the different partition;

executing in a third state of pre-copy, wherein available memory pages of the virtual machine are copied from the virtual machine to the newly initialized virtual machine;

executing in a fourth state of stop-and-copy, wherein the virtual machine stops execution and any memory pages of the virtual machine that are out of synchronization with the memory pages of the newly initialized virtual machine are copied and/or re-copied to the newly initialized virtual machine;

executing in a fifth state of commitment, wherein the virtual machine corresponding to the first partition is eliminated and the newly initialized virtual machine is then referred to as the virtual machine; and

executing in a sixth state of activation, wherein the virtual machine connects to local devices and resumes normal operation by entering into a first state of pre-migration executing from the different partition.

13. The method of claim 9, further comprising:

selecting the different partition from among a plurality of available partitions within the computing apparatus based on a comparison of load statuses for each of the respective plurality of available partitions.

14. The method of claim 9, further comprising:

selecting the different partition from among a plurality of available partitions within the computing apparatus based on a comparison of VM processor core temperatures for each of the respective plurality of available partitions.

15. The method of claim 9, wherein the computing apparatus comprises:

a first virtual machine monitor to manage the one or more virtual machines to execute on the first partition; and

a second virtual machine monitor to manage the one or more virtual machines to execute on the different partition.

16. A computer-readable storage medium having instructions stored thereon that, when executed by a plurality of processor cores within a single computing apparatus, cause the computing apparatus to perform a method comprising:

hosting a plurality of partitions on the computing apparatus, wherein each of the plurality of partitions is allocated a subset of processor cores within the computing apparatus without overlap or double assignment of any of the processor cores to more than one of the plurality of partitions;

executing one or more virtual machines within each of the plurality of partitions of the computing apparatus, wherein each virtual machine executes via the subset of processor cores assigned to the respective partition in which the one or more virtual machines operate;

determining a load status of the subset of processor cores allocated to a first partition among the plurality of partitions within the computing apparatus and comparing the load status of the first partition against an upper threshold for the first partition;

identifying a virtual machine among the one or more virtual machines executing on the subset of processor cores allocated to the first partition which is running a task not present on a native task list for the first partition; and

migrating the virtual machine running the task which is not present on the native task list for the first partition to a different partition within the computing apparatus when the load status of the first partition exceeds the upper threshold for the first partition.

17. The computer-readable storage medium of claim 16:

wherein the first partition comprises a system partition to perform system tasks initiated on behalf of an end-user, rather than initiated directly by an end-user of the computing apparatus; and

wherein migrating the virtual machine to a different partition within the computing apparatus comprises migrating the virtual machine to a user partition that performs general computing tasks initiated directly by the end-user.

18. The computer-readable storage medium of claim 16, wherein the method further comprises selecting the different partition from among a plurality of available partitions within the computing apparatus based on a comparison of VM processor core temperatures for each of the respective subset of processor cores allocated to each of the plurality of available partitions.

19. The computer-readable storage medium of claim 16, wherein the computing apparatus comprises:

a virtual machine monitor to manage the one or more virtual machines executing in the first partition and the one or more virtual machines executing in the different partition.

20. The computer-readable storage medium of claim 16, wherein the computing apparatus comprises:

a first virtual machine monitor to manage the one or more virtual machines to execute on the first partition; and

a second virtual machine monitor to manage a second one or more virtual machines to execute on the different partition, wherein the first virtual machine monitor and the second virtual machine monitor are connected via a communication channel having a shared memory component, and/or a direct memory access component to allow the first and second virtual machine monitors to exchange data corresponding to the first partition and the different partition.