Optimizing virtual machines placement in cloud computing environments

• US 10,719,343 B2
• Filed: 10/14/2016
• Issued: 07/21/2020
• Est. Priority Date: 10/12/2011
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Podvratnik, Amadeus, Spatzier, Thomas, Teich, Torsten
• Primary Examiner(s): Bullock, Jr., Lewis A
• Assistant Examiner(s): Kepnang, Gilles R
• Application Number:
  US15/294,550
• Publication Number:
  US 20170031706A1
• Time in Patent Office: 1,376 Days
• CPC Class Codes: G06F 2009/45562 Creating, deleting, cloning virtual machine instances
• CPC Class Codes: G06F 2009/45591 Monitoring or debugging support
• CPC Class Codes: G06F 2209/5019 Workload prediction
• CPC Class Codes: G06F 8/61 Installation
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/5027 the resource being a machine, e.g. CPUs, Servers, Terminals
• CPC Class Codes: G06F 9/5072 Grid computing
• CPC Class Codes: G06F 9/5077 Logical partitioning of resources; Management or configuration of virtualized resources specific details on emulation or internal functioning of virtual machines G06F9/455

Abstract

Placement of instances of virtual machine (VM) images in a cloud computing environment is optimized. A request is received to deploy a new VM image instance. VM resource usage profiles are read that indicate predicted resource consumption data of the VM image. Resource pool profiles for multiple resource pools are read that indicate predicted resource consumption data for the resource pools. For each of the resource pools, the VM resource usage profile is aligned with the respective resource pool profile. An aggregated resource consumption profile is calculated as an aggregation of the predicted resource consumptions of the aligned profiles. A determination is made of one of multiple resource pools with the least predicted aggregated resource consumption, and the VM image is instantiated on the determined one of the multiple resource pools with the least predicted aggregated resource consumption over time.

Claims

1. A method, comprising:

by a placement engine that instantiates virtual machine (VM) instances;

receiving a request from a user to deploy an instance of a VM image;

reading a VM resource usage profile (RUP) that indicates a predicted resource consumption of the requested instance of the VM image over a first period of time;

retrieving an existing user-specific VM resource usage profile (URUP) for the instance of the VM image requested to be deployed over the first period of time;

determining user-specific patterns based on combined respective data of the RUP and respective data of the URUP;

reading one resource pool profile for each of a plurality of resources pools, where each resource pool profile indicates a predicted resource consumption in the respective resource pool over a second period of time;

for each resource pool;

generating a resulting VM resource usage profile from a selected one of the VM RUP and the URUP;

analyzing the resulting VM resource usage profile and the resource pool profile, the analysis comprising aligning the resulting VM resource usage profile and the respective resource pool profile of the respective resource pool by aligning the second period of time of the respective resource pool profile with the first period of time of the resulting VM resource usage profile; and

calculating, for existing host platforms, an aggregated resource consumption profile that predicts a total resource consumption over time in the resource pool as an aggregation of the predicted resource consumptions of the aligned profiles by comparing the resulting VM resource usage profile with the resource pool profile of the respective resource pool, wherein the resource pool profile of the respective resource pool is obtained as a superposition of an assigned resource usage profile of all instances of all VMs running on the resource pool;

determining, from the aggregated resource consumption profiles, a resource pool of the plurality of resource pools with a least predicted aggregated resource consumption over time;

determining that the resource pool with the least predicted aggregated resource consumption over time has no free resource on the existing host platform to fit the instance of the VM image, based on a current allocation of resources;

calculating a resource shortage for the determined least predicted aggregated resource pool;

based on the calculated resource shortage, performing, by a resource manager, a reallocation of at least one resource, initially allocated to an instantiated VM image instance in the determined least predicted aggregated resource pool, to the requested VM image, wherein the aggregated resource consumption profiles are used as input for determining another instantiated VM image instance to which the at least one resource is to be reallocated; and

subsequent to reallocating the at least one resource, instantiating the VM image on the least predicted resource pool expected to provide the best available fit for the resource requirements of the VM image.

2. The method of claim 1, where:

if a user-specific VM resource usage profile associated with the user is available and further indicates a predicted user-specific resource consumption of the requested instance of the VM image by the user over the first period of time, where the aligning and calculating are performed using the user-specific VM resource usage profile instead of the VM resource usage profile.

3. The method of claim 2, where:

the user-specific VM resource usage profile comprises user-specific conditions on at least one of a number, a type, and an availability of resources for guaranteeing that at least one of the number, the type, and the availability of the resources allocated to the requested instance of the VM image is in accordance with a service level agreement (SLA); and

determining, from the aggregated resource consumption profiles, the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time comprises;

determining, from the aggregated resource consumption profiles, the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time that additionally satisfies the user-specific conditions.

4. The method of claim 1, where the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time is selected from a group consisting of:

a resource pool for which a lowest integral resource consumption value of the aggregated resource consumption profile was calculated; and

a resource pool for which the lowest integral resource consumption value of the aggregated resource consumption profile was calculated and where the aggregated resource consumption profile does not comprise a value exceeding a resource consumption threshold value.

5. The method of claim 1, further comprising:

evaluating, by a VM usage supervisor module, a current resource consumption of a plurality of instances of the VM image to provide first evaluated data;

predicting resource consumption of any instance of the VM image over the first period of time by evaluating the first evaluated data; and

one of generating and updating the VM resource usage profile of the VM image using the predicted resource consumption.

6. The method of claim 1, further comprising:

monitoring a number and a type of the resources of the resource pool allocated to a plurality of VM image instances running on the resource pool to provide monitoring data, the monitoring data comprising a measured physical assignment of resources to the plurality of VM image instances over the second period of time;

superposing the monitoring data obtained for the plurality of VM image instances running on the resource pool to generate a superposed resource consumption profile for the resource pool; and

using the superposed resource consumption profile to one of generate and update the resource pool profile of the resource pool.

7. The method of claim 1, where each VM image instance running on the resource pool has one or more assigned resources of the resource pool, further comprising:

determining, by a VM usage supervisor, all resources in the resource pool that are unoccupied over the second period of time by subtracting the aggregated resource consumption profile from a known number of all resources of the resource pool;

determining, for the requested instance of the VM image instantiated on the resource pool, one or more currently required resources by evaluating one of the VM resource usage profile and a user-specific VM resource usage profile of the VM image; and

triggering, by the placement engine, the instantiation of the requested instance of the VM image on another resource pool, if a number of resources of a first resource type determined to be required by the requested instance of the VM image exceeds the determined number of currently unoccupied resources of the first resource type in the resource pool.

8. The method of claim 1, further comprising performing, by a resource manager, a reallocation of at least one resource initially allocated to the instantiated VM image instance, where the aggregated resource consumption profiles are used as input for determining another instantiated VM image instance to which the at least one resource is to be reallocated.

9. The method of claim 8, further comprising updating at least one resource pool profile of the plurality of resource pools with new resource consumption data resulting from the reallocation of the at least one resource or from a movement of the instantiated VM image instance.

10. The method of claim 1, where one of the predicted resource consumption indicated in the VM resource usage profile and predicted resource usage data specified in a user-specific VM resource usage profile is used to determine if resources required by the requested instance of the VM image in one or more future moments of time will be available in one of the plurality of resource pools in the one or more future moments in time.

11. A system, comprising:

a storage device comprising a plurality of virtual machine (VM) images;

a plurality of resource pools, each resource pool comprising one or more resources;

an interface; and

a placement engine device programmed to;

receive, via the interface, a request from a user to deploy an instance of a VM image;

read, from the storage device, a VM resource usage profile (RUP) that indicates a predicted resource consumption of the requested instance of the VM image over a first period of time;

retrieve, from the storage device, a user-specific VM resource usage profile (URUP) for the instance of the VM image requested to be deployed over the first period of time;

determine user-specific patterns based on combined respective data of the RUP and respective data of the URUP;

read, from the storage device, one resource pool profile for each of the plurality of resources pools, where each resource pool profile indicates a predicted resource consumption in the respective resource pool over a second period of time;

for each resource pool;

generate a resulting VM resource usage profile from a selected one of the VM RUP and the URUP;

analyze the resulting VM resource usage profile and the resource pool profile, the analysis comprising a step to align the resulting VM resource usage profile and the respective resource pool profile of the respective resource pool by aligning the second period of time of the respective resource pool profile with the first period of time of the resulting VM resource usage profile; and

calculate, for existing host platforms, an aggregated resource consumption profile that predicts a total resource consumption over time in the resource pool as an aggregation of the predicted resource consumptions of the aligned profiles by comparing the resulting VM resource usage profile with the resource pool profile of the respective resource pool, wherein the VM RUP of the respective resource pool is obtained as a superposition of an assigned resource usage profile of all instances of all VMs running on the resource pool;

determine, from the aggregated resource consumption profiles, a resource pool of the plurality of resource pools with a least predicted aggregated resource consumption over time;

determine that the resource pool with the least predicted aggregated resource consumption over time has no free resource on the existing host platforms to fit the instance of the VM image, based on a current allocation of resources;

calculate a resource shortage for the determined least predicted aggregated resource pool;

and based on the calculated resource shortage, perform, by a resource manager, a reallocation of at least one resource, initially allocated to an instantiated VM image instance in the determined least predicted aggregated resource pool, to the requested VM image, wherein the aggregated resource consumption profiles are used as input for determining another instantiated VM image instance to which the at least one resource is to be reallocated; and

subsequent to reallocating the at least one resource, instantiate the VM image on the least predicted resource pool expected to provide the best available fit for the resource requirements of the VM image.

12. The system of claim 11, where:

if a user-specific VM resource usage profile associated with the user is available and further indicates a predicted user-specific resource consumption of the requested instance of the VM image by the user over the first period of time, where the aligning and calculating are performed using the user-specific VM resource usage profile instead of the VM resource usage profile.

13. The system of claim 11, where the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time is selected from a group consisting of:

a resource pool for which a lowest integral resource consumption value of the aggregated resource consumption profile was calculated; and

a resource pool for which the lowest integral resource consumption value of the aggregated resource consumption profile was calculated and where the aggregated resource consumption profile does not comprise a value exceeding a resource consumption threshold value.

14. A computer program product, comprising:

a computer readable storage memory having computer readable program code embodied therewith, where the computer readable program code when executed on a computer causes the computer to;

receive a request from a user to deploy an instance of a virtual machine (VM) image;

read a VM resource usage profile (RUP) that indicates a predicted resource consumption of the requested instance of the VM image over a first period of time;

retrieve a user-specific VM resource usage profile (URUP) for the instance of the VM image requested to be deployed over the first period of time;

determine user-specific patterns based on combined respective data of the RUP and respective data of the URUP;

read one resource pool profile for each of the plurality of resources pools, where each resource pool specific profile indicates a predicted resource consumption in the respective resource pool over a second period of time;

for each resource pool;

generate a resulting VM resource usage profile from a selected one of the VM RUP and the URUP;

analyze the resulting VM resource usage profile and the resource pool profile, the analysis comprising a step to align the resulting VM resource usage profile and the respective resource pool profile of the respective resource pool by aligning the second period of time of the respective resource pool profile with the first period of time of the resulting VM resource usage profile; and

calculate, for existing host platform, an aggregated resource consumption profile that predicts a total resource consumption over time in the resource pool as an aggregation of the predicted resource consumptions of the aligned profiles by comparing the resulting VM resource usage profile with the resource pool profile of the respective resource pool, wherein the VM RUP of the respective resource pool is obtained as a superposition of an assigned resource usage profile of all instances of all VMs running on the resource pool;

determine, from the aggregated resource consumption profiles, a resource pool of the plurality of resource pools with a least predicted aggregated resource consumption over time;

determine that the resource pool with the least predicted aggregated resource consumption over time has no free resource on the existing host platforms to fit the instance of the VM image, based on a current allocation of resources;

calculate a resource shortage for the determined least predicted aggregated resource pool;

based on the calculated resource shortage, perform, by a resource manager, a reallocation of at least one resource, initially allocated to an instantiated VM image instance in the determined least predicted aggregated resource pool, to the requested VM image, wherein the aggregated resource consumption profiles are used as input for determining another instantiated VM image instance to which the at least one resource is to be reallocated; and

subsequent to reallocating the at least one resource, instantiate the VM image on the last predicted resource pool expected to provide the best available fit for the resource requirements of the VM image.

15. The computer program product of claim 14, where:

if a user-specific VM resource usage profile associated with the user is available and further indicates a predicted user-specific resource consumption of the requested instance of the VM image by the user over the first period of time, where the computer readable program code when executed on the computer further causes the computer to perform the aligning and calculating using the user-specific VM resource usage profile instead of the VM resource usage profile.

16. The computer program product of claim 15, where:

the user-specific VM resource usage profile comprises user-specific conditions on at least one of a number, a type, and an availability of resources for guaranteeing that at least one of the number, the type, and the availability of the resources allocated to the requested instance of the VM image is in accordance with a service level agreement (SLA); and

where, in causing the computer to determine, from the aggregated resource consumption profiles, the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time, the computer readable program code when executed on the computer causes the computer to;

determine, from the aggregated resource consumption profiles, the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time that additionally satisfies the user-specific conditions.

17. The computer program product of claim 14, where the resource pool of the plurality of resource pools with the least predicted aggregated resource consumption over time is selected from a group consisting of:

a resource pool for which a lowest integral resource consumption value of the aggregated resource consumption profile was calculated; and

a resource pool for which the lowest integral resource consumption value of the aggregated resource consumption profile was calculated and where the aggregated resource consumption profile does not comprise a value exceeding a resource consumption threshold value.

18. The computer program product of claim 14, where the computer readable program code when executed on the computer further causes the computer to:

monitor a number and a type of the resources of the resource pool allocated to a plurality of VM image instances running on the resource pool to provide monitoring data, the monitoring data comprising a measured physical assignment of resources to the plurality of VM image instances over the second period of time;

superpose the monitoring data obtained for the plurality of VM image instances running on the resource pool to generate a superposed resource consumption profile for the resource pool; and

use the superposed resource consumption profile to one of generate and update the resource pool profile of the resource pool.