Techniques for workload balancing among a plurality of physical machines

• US 9,489,222 B2
• Filed: 08/21/2012
• Issued: 11/08/2016
• Est. Priority Date: 08/24/2011
• Status: Active Grant

• Current Assignee: Radware Limited
• Original Assignee: Radware Limited
• Inventors: Zlotkin, Gilad
• Primary Examiner(s): Puente, Emerson
• Assistant Examiner(s): Kamran, Mehran
• Application Number:
  US13/590,476
• Publication Number:
  US 20130055260A1
• Time in Patent Office: 1,540 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 9/45533 Hypervisors; Virtual machine monitors
• CPC Class Codes: G06F 9/4856 resumption being on a different machine, e.g. task migration, virtual machine migration G06F9/5088 takes precedence

Abstract

A method for workload balancing among a plurality of physical machines hosting a plurality of virtual machines (VMs) is disclosed. The method comprises periodically measuring a utilization of each hardware resource in each of the plurality of physical machines; computing a resource utilization score for each hardware resource based on its respective measured utilization; computing a total physical machine utilization score for each physical machine based on the computed resource utilization scores of its respective resources; and upon reception of a client request corresponding to a software application, selecting one physical machine of the plurality of physical machines to serve the client request, wherein the selection is based on the computed total physical machine utilization.

Claims

1. A method for workload balancing among a plurality of physical machines hosting a plurality of virtual machines (VMs), comprising:

periodically measuring a utilization of each hardware resource in each of the plurality of physical machines that host VMs that can execute an application to be load balanced;

computing a resource utilization score for each hardware resource based on its type, respective measured utilization, weight factor and load threshold;

wherein the load threshold is a predefined threshold set for a specific hardware resource type;

wherein the weight factor is determined based on the relevancy of the measured resource utilization for a proper execution of the application;

computing a total physical machine utilization score for each physical machine based on the computed resource utilization scores of its respective resources;

upon reception of a client request corresponding to a software application, selecting one physical machine of the plurality of physical machines to serve the client request, wherein the selection is based on the computed total physical machine utilization; and

forwarding the client request to a VM hosted on the selected physical machine, wherein all of the plurality of VMs hosted on the plurality of physical machines run an instance of the software application.

2. The method of claim 1, wherein the physical machine with a lowest total physical machine utilization score is selected to serve the client request.

3. The method of claim 2, wherein each hardware resource is any one of:

one or more central processing unit (CPU) cores, physical memory, input/output interfaces, storage, and one or more network interface cards (NICs).

4. The method of claim 3, wherein a measured utilization of a hardware resource is any one of:

utilization of the one or more CPU cores, a percentage allocated memory out of the total available physical memory, a percentage of a bandwidth utilization out of a total available bandwidth on the one or more NICs, a input/output (I/O) operations rate, and a memory page swap rate (MPR).

5. The method of claim 4, wherein the measured utilization of each hardware resource is an average of a plurality of utilization measurements for the respective hardware resource over time.

6. The method of claim 4, wherein the resource utilization score for each hardware resource s computed as follows:

weight(r)*(1−

resource_utilization(pm,r))/load threshold(r)),

wherein resource_utilization(pm, r) is the measured resource utilization for the hardware resource, and load_threshold is a predefined threshold set for a specific hardware resource type.

7. The method of claim 6, wherein the weight (r) factor is determined based on the relevancy of the measured resource utilization for a proper execution of the software application.

8. The method of claim 7, wherein the total physical machine utilization score of a physical machine is a sum of utilization resource scores computed for the hardware resources of the physical machine divided by a sum of the weight factors of the hardware resources.

9. The method of claim 1, further comprising:

computing a load-balancing weight for each physical machine of the plurality of physical machines; and

distributing client'"'"'s requests to the plurality of physical machines proportional to a computed load-balancing weight of each physical machine of the plurality of physical machines.

10. The method of claim 9, wherein each physical machine of the plurality of physical machines hosts at least one VM configured to run an instance of the software application.

11. The method of claim 10, wherein the plurality of physical machines are operable in a plurality of datacenters, wherein the client'"'"'s requests are forwarded to a group of physical machines connected in a datacenter by means of an application delivery controller (ADC), wherein the method is performed by the ADC connected in each datacenter.

12. The method of claim 11, wherein a plurality of ADCs connected in the plurality of datacenters communicate with an advisory server, wherein the advisory server selects the physical machine.

13. A non-transitory computer readable medium having stored thereon instructions for causing a processor to perform a method of claim 1.

14. A method for workload balancing among a plurality of physical machines hosting a plurality of virtual machines (VMs), comprising:

periodically measuring a utilization of each hardware resource in each of the plurality of physical machines that host VMs that can execute an application to be load balanced;

computing a resource utilization score for each hardware resource based on its type, respective measured utilization, weight factor and load threshold;

wherein the load threshold is a predefined threshold set for a specific hardware resource type;

wherein the weight factor is determined based on the relevancy of the measured resource utilization for a proper execution of the application;

computing a total physical machine utilization score for each physical machine based on the computed resource utilization scores of its respective resources;

computing a load-balancing weight for each physical machine based on its computed total physical machine utilization score; and

distributing client requests corresponding to a software application to the plurality of physical machines proportional to the load-balancing weight computed for each physical machine, wherein the client requests are served by the plurality of VMs hosted on the plurality of physical machines.

15. The method of claim 14, wherein the load-balancing weight for each physical machine is computed as a round number of a predefined value divided by the total physical machine utilization score calculated for the respective physical machine.

16. The method of claim 14, wherein all of the plurality of VMs hosted on the plurality of physical machines run an instance of the software application, and wherein each physical machine of the plurality of physical machines hosts at least one VM configured to run an instance of the software application.

17. A non-transitory computer readable medium having stored thereon instructions for causing a processor to perform a method of claim 14.

18. A system of workload balancing among a plurality of physical machines hosting a plurality of virtual machines (VMs), comprising:

a memory for storing executable instructions;

a processor connected to a memory and operable to execute instructions in the memory to configure the system to;

periodically measure a utilization of each hardware resource in each of the plurality of physical machines that host VMs that can execute an application to be load balanced;

compute a resource utilization score for each hardware resource based on its type, respective measured utilization, weight factor and load threshold;

wherein the load threshold is a predefined threshold set for a specific hardware resource type;

wherein the weight factor is determined based on the relevancy of the measured resource utilization for a proper execution of the application;

compute a total physical machine utilization score for each physical machine based on the computed resource utilization scores of its respective resources, wherein the total physical machine utilization score of a physical machine is a sum of utilization resource scores computed for the hardware resources of the physical machine divided by a sum of the weight factors of the hardware resources;

select, based on the computed total physical machine utilization score, one physical machine from the plurality of physical machines to serve an incoming client request related to an execution of a software application; and

forward the client request to a VM hosted on the selected physical machine, wherein all of the plurality of VMs hosted on the plurality of physical machines run an instance of the software application.

19. The system of claim 18, wherein each physical machine of the plurality of physical machines hosts at least one VM configured to run an instance of the software application.

20. The system of claim 19, wherein the system is integrated in any one of:

an application delivery controller (ADC) and an advisory server.

21. The system of claim 20, wherein the plurality of physical machines are operable in a plurality of datacenters, wherein client'"'"'s requests are forwarded to a group of physical machines connected in a datacenter by means of ADC, wherein ADCs connected in the plurality of datacenters communicate with an advisory server.