KR101924467B1 - System and method of resource allocation scheme for cpu and block i/o performance guarantee of virtual machine - Google Patents

Abstract

The present invention relates to a virtual technology and, more specifically, to a resource management scheduler in a virtual system, in which a resource allocation system according to one embodiment can comprise a statistics collector for collecting performance information of a virtual machine including input/output operations per second (IOPS), I/O bandwidth and latency and CPU utilization, an SLA manager which calculates a service level agreement (SLA) satisfaction required by each virtual machine based on performance information of the collected virtual machines, and a differential distribution processing unit for controlling the differential distribution of the CPU resource and the block I/O resource to each of the virtual machines based on the calculated SLA satisfaction.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a resource allocation system and method for guaranteeing performance in CPU and block I / O operations of a virtual machine,

The present invention relates to virtualization technology, and more particularly, to a resource management scheduler in a virtualization system.

Kernel-based virtual machine (KVM) is a virtualization platform that uses Linux, the host operating system, as a hypervisor. The virtualization environment is configured through the kernel module of Linux, the host operating system, and QEMU, an open source based machine emulator, is used for virtualizing I / O devices. By providing a virtual machine with a para-virtualization I / O device, you can reduce the overhead of emulation for virtualization. A virtual machine is configured in the form of a single process, and a virtual CPU (vCPU) or a virtual I / O device provided to the virtual machine is configured in the form of a thread. Therefore, the performance of the virtual machine is determined according to a method of scheduling the virtual devices to be provided to the virtual machine.

The Complete Fair Scheduler (CFS) is a CPU scheduler that is used by default in Linux. KVM, a virtualization platform that uses Linux as the host operating system, is used to schedule vCPUs or virtual I / O devices to provide to virtual machines. The basic policy of CFS is to allocate fair resources to all processes. However, resource allocation can be differentiated when a process-specific weight is set using a control group (Cgroup). In addition, an absolute CPU occupancy rate can be determined by setting a time period during which a process can occupy the CPU in a predetermined cycle. This prevents a process from monopolizing the CPU.

Virtualization technology is a technology that provides an environment that can increase resource utilization of physical machines by allowing many virtual machines to share resources of physical machines. It is a technology that provides the platform Technology.

Because of the nature of the cloud that provides the service to the user, ensuring the performance required by the user is an important element of the cloud. Therefore, cloud providers and users enter into a service level agreement (SLA) and contracts for compensation in case of violation.

However, resource sharing, which is a characteristic of virtualization technology, may cause resource competition between virtual machines, which may result in performance interference. Performance interference is a major factor that makes it difficult to guarantee the performance required by users in an environment like a cloud. In order to solve this problem, it is necessary to efficiently allocate resources to satisfy the required performance of each virtual machine. However, in a virtualized environment, the hypervisor requires an emulation process for I / O processing and emulation requires CPU resources. Therefore, I / O resources as well as CPU resources must be considered to ensure I / O operations. Therefore, in a virtualization environment in which different tasks are performed, it is necessary to consider various resources in order to guarantee the required performance of each task.

Korean Patent No. 1070431 "Virtualization-based physical system and its resource management method" Korean Patent No. 1277273 "Resource allocation method between terminal devices using resource management system and resource management server therefor"

The present invention aims at ensuring each performance requirement of a virtual machine performing a CPU operation and a virtual machine performing a block I / O operation by considering a CPU and a block I / O resource in a KVM-based virtualization environment do.

An object of the present invention is to set a required performance of a block I / O or a CPU of a virtual machine in a hypervisor and to distribute the CPU and block I / O resources with the same standard to achieve required performance of a virtual machine.

The present invention aims at satisfying required performance of a virtual machine for various requirements and situations because it distributes a CPU and a block I / O resource differentially to a virtual machine through feedback control.

A resource allocation system according to an embodiment of the present invention includes a statistics collector for collecting performance information of a virtual machine including IOPS, I / O bandwidth, delay time, and CPU utilization rate, An SLA manager for calculating a SLA (Service Level Agreement) satisfaction required by each virtual machine, and a differential distribution processor for controlling to distribute the CPU resources and the block I / O resources to the respective virtual machines based on the calculated SLA satisfaction . ≪ / RTI >

The resource allocation method according to an exemplary embodiment of the present invention includes collecting performance information of a virtual machine including IOPS, I / O bandwidth, delay time, and CPU utilization rate in a statistic collector. In the SLA manager, Calculating SLA (Service Level Agreement) satisfaction required by each virtual machine based on the performance information of the virtual machine, and calculating a CPU resource and a block I / O to each of the virtual machines based on the calculated SLA satisfaction, And controlling to distribute the resources differentially.

According to the present invention, in consideration of a combination of CPU and block I / O resources in a KVM-based virtualization environment, each performance requirement of a virtual machine performing a CPU operation and a virtual machine performing a block I / O operation can be guaranteed have.

In addition, the performance requirements of the virtual machine can be achieved by setting the block I / O of the virtual machine and the performance requirements of the CPU to the hypervisor, and distributing the CPU and the block I / O resources with the same reference.

In addition, since the CPU and block I / O resources are differentially distributed to the virtual machine through the feedback control, the required performance of the virtual machine can be satisfied for various requirements and situations.

1 is a functional block diagram of a resource allocation system according to an embodiment of the present invention.
2 is a block diagram illustrating the resource allocation system shown in FIG. 1 in more detail.
3 is a flowchart illustrating a resource allocation method according to an embodiment of the present invention.
4 is a diagram showing SLA satisfaction with respect to CPU SLA and I / O SLA by time.
FIG. 5 is a diagram showing SLA satisfaction when the GoS framework is used. FIG.

It is to be understood that the specific structural or functional descriptions of embodiments of the present invention disclosed herein are only for the purpose of illustrating embodiments of the inventive concept, But may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein.

Embodiments in accordance with the concepts of the present invention are capable of various modifications and may take various forms, so that the embodiments are illustrated in the drawings and described in detail herein. However, it is not intended to limit the embodiments according to the concepts of the present invention to the specific disclosure forms, but includes changes, equivalents, or alternatives falling within the spirit and scope of the present invention.

The terms first, second, or the like may be used to describe various elements, but the elements should not be limited by the terms. The terms may be named for the purpose of distinguishing one element from another, for example without departing from the scope of the right according to the concept of the present invention, the first element being referred to as the second element, Similarly, the second component may also be referred to as the first component.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between. Expressions that describe the relationship between components, for example, "between" and "immediately" or "directly adjacent to" should be interpreted as well.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises ", or" having ", and the like, are used to specify one or more of the features, numbers, steps, operations, elements, But do not preclude the presence or addition of steps, operations, elements, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the meaning of the context in the relevant art and, unless explicitly defined herein, are to be interpreted as ideal or overly formal Do not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the scope of rights of the present invention is not limited or limited by these embodiments. In addition, the same reference numerals shown in the drawings denote the same members.

1 is a functional block diagram of a resource allocation system according to an embodiment of the present invention.

Virtualization guarantees independent space for each virtual machine on the physical machine. At this time, the hypervisor provides the virtual machine with a vCPU, which is a virtual CPU, to the virtual machine, and the virtual machine executes the process of the virtual machine within the vCPU process.

In addition, I / O requests originating from the virtual machine are delivered to the real device through I / O emulation in the hypervisor. I / O emulation is the implementation of the virtual I / O devices provided to the virtual machine by software rather than the actual device. Thus, I / O emulation is software and requires CPU resources to operate. Therefore, there is a need to consider the CPU resource allocation in order to guarantee the performance of the I / O operation.

The resource allocation system 100, which may be referred to as a resource allocation device, a resource allocation server, and the like, takes a complex consideration of CPU and block I / O resources in a KVM-based virtualization environment, You can ensure each performance requirement of the virtual machine performing the work.

To this end, the resource allocation system 100 may include a statistics collector 110, an SLA manager 120, and a differential distribution processing unit 130.

The statistics collector 110 may collect performance information of the virtual machine including input / output operations per second (IOPS), I / O bandwidth, and latency and CPU utilization.

Next, the SLA manager 120 can calculate the SLA (Service Level Agreement) satisfaction required by each virtual machine based on the collected performance information of the virtual machine. For example, the SLA manager 120 can manage the SLA value, the SLA satisfaction, the resource usage rate used by the virtual machine, and the performance information of the virtual machine. In addition, the SLA manager 120 may provide an API for setting the SLA value and confirming the SLA satisfaction.

The differential distribution processor 130 may control to distribute the CPU resource and the block I / O resource to each virtual machine on the basis of the calculated SLA satisfaction. The differential distribution processor 130 according to an embodiment may deliver the calculated SLA satisfaction to the CPU scheduler or the I / O scheduler so as to control the distribution of the CPU resources and the block I / O resources to each virtual machine.

More specifically, the CPU scheduler can perform the resource differential distribution by utilizing the functions provided by the Linux Complete Fair Scheduler (CFS).

In addition, the CPU scheduler according to an embodiment performs a resource differential distribution function, and divides virtual machines operating in the physical machine into a plurality of groups. The SLA unsatisfied group, the SLA A supersaturated group, and an SLA satisfying group, and calculates a quota value of each classified group, and performs resource differentiation distribution considering the calculated quota value of each group. The quota value represents the CPU time available to each group, and performs a CPU resource differential distribution by adjusting the quota value.

The CPU scheduler according to an embodiment can calculate a quota value required for reaching 100% of SLA satisfaction of virtual machines in an SLA unsatisfied group through feedback control.

Also, the CPU scheduler can perform feedback control of the quota using Equation (1).

[Equation 1]

In addition, the CPU scheduler according to an exemplary embodiment may calculate a quota value that can be transferred to virtual machines of the SLA unsatisfied group by virtual machines belonging to the SLA supersaturated group, and may determine that each virtual machine in the SLA superset group You can calculate the value of the quota to decrease proportionally to reach 100%.

Meanwhile, the CPU scheduler according to the embodiment compares values calculated for the SLA unsatisfied group and each of the virtual machines of the supersaturated group by groups, compares the two values, and assigns a smaller value among the comparison results to the quota It is also possible to select the value for exchange.

The CPU scheduler according to an exemplary embodiment may divide the value of Quota in the SLA unsatisfied group according to the SLA satisfaction rate of the virtual machines belonging to the group and decrease according to the SLA satisfaction ratio in the supersaturation group.

On the other hand, the I / O scheduler can modify the Opportunistic I / O Scheduler (OIOS) technique to feedback control according to SLA satisfaction. In addition, the I / O scheduler may schedule a DAT (Differentially Allowed Time) value indicating a time when a virtual machine can transmit an I / O request to a block I / O device so as to satisfy the required performance of the virtual machine .

2 is a block diagram illustrating the resource allocation system shown in FIG. 1 in more detail.

The resource allocation system 200 includes a structure of an integrated resource allocation scheme of a CPU and a block I / O resource to guarantee the required performance of a CPU or a block I / O task running in the proposed virtual machine.

The SLA manager (220) manages the performance requirements of the user, SLA value and SLA satisfaction, resource utilization rate used by the virtual machine, and performance information of the virtual machine. It also provides system administrators with APIs for setting SLA values and checking SLA satisfaction, and allowing the administrator to control the framework.

For example, the SLA value can be set to a value related to the CPU and block I / O, and the CPU SLA can set the number of CPUs that the virtual machine always wants to occupy independently of the number of virtual machines' vCPUs.

The I / O SLA sets the block I / O bandwidth required by the virtual machine to a value. Then, the SLA satisfaction is calculated based on the performance information and the resource utilization rate of the virtual machine, and the value is transferred to each resource scheduler so that the CPU and the block I / O resources are distributed to the virtual machines based on the SLA satisfaction.

In addition, the information necessary for calculating the SLA satisfaction can be collected by the statistic collector 210. The statistics collector 210 may collect IOPS, I / O bandwidth, latency, and CPU usage information of the virtual machine's block I / O performance information.

The CPU scheduler 230 and the I / O scheduler 240 represent a CPU and a block I / O resource allocation scheduler. Each resource scheduler distributes the resources to the virtual machine in such a way that the satisfaction degree reaches or exceeds 100% based on the SLA satisfaction received from the SLA manager 220.

As the CPU scheduler 230, there is a component that enables resource differential distribution by utilizing the functions provided by CFS based on CFS, which is a basic scheduler of Linux. The component can use the quota function provided by CFS.

Quota is a value indicating the time to occupy the physical CPU in a certain period, and is a function that allows setting of the actual usage time value for the CPU.

The present invention defines a corresponding component as a CPU resource differential distribution component, and the component adjusts a quota value through a feedback control function. That is, feedback control is performed on the amount of the quota required for the satisfaction of the SLA received from the SLA manager 220 to exceed 100%.

The I / O scheduler 240 modifies the existing opportunistic I / O scheduler (OIOS) technique to feedback control according to SLA satisfaction. OIOS is a technique that satisfies the required performance of a virtual machine by feedback control of a DAT (Differentially Allowed Time) value indicating a time for a virtual machine to transmit an I / O request to a block I / O device.

The detailed operation of the CPU resource differential distribution component can divide the virtual machines operating in the physical machine into three groups.

Specifically, the group with less than 99% SLA satisfaction is classified as an SLA unsatisfactory group, the group with more than 101% SLA supersaturation group, and the group with 99 ~ 101% SLA satisfaction group can do. And, the feedback control is used to calculate the quota value required for the SLA satisfaction of the virtual machines in the SLA unsatisfied group to reach 100%.

The feedback control method calculates a quota value required to reach 100% SLA satisfaction in the current state based on the ratio of increase in quota and increase in SLA satisfaction measured in the previous feedback period.

In addition, the virtual machines in the SLA supersaturation group can calculate quota values that can be transferred to the virtual machines in the SLA dissatisfied group. That is, each virtual machine in the SLA supersaturated group computes the quota of the quota to be reduced to reach 100% of the current SLA satisfaction.

The values calculated for each virtual machine in the SLA-unsatisfied group and the supersaturated group are grouped by group and compared. The smaller value is selected as the value for the quota exchange between the unsatisfied group and the supersaturated group. In the unsatisfied group, the value is distributed according to the SLA satisfaction ratio of the virtual machines belonging to the group, and in the supersaturation group, it can be reduced according to the SLA satisfaction ratio. This allows the virtual machine of the SLA dissatisfied group to enter the SLA satisfaction group and prevents the virtual machine of the SLA superset group from moving to the SLA dissatisfied group.

In the above-described manner, the present invention considers that a block I / O task has a dependence on a CPU and a block I / O resource, and allocates CPU resources and block I / O resources to a virtual machine based on SLA satisfaction To satisfy the CPU or block I / O performance required by each virtual machine.

3 is a flowchart illustrating a resource allocation method according to an embodiment of the present invention. The resource allocation method may be performed in the resource allocation system shown in FIG. 1 or FIG.

A resource allocation method according to an exemplary embodiment may collect performance information of a virtual machine (step 301).

For example, the resource allocation method may collect performance information of a virtual machine including IOPS, I / O bandwidth, and latency and CPU utilization.

Next, the resource allocation method can calculate SLA (Service Level Agreement) satisfaction required by each virtual machine (step 302). For example, based on the collected performance information of the virtual machine, the SLA satisfaction required by each virtual machine can be calculated.

In addition, the resource allocation method can control to distribute the CPU resource and the block I / O resources to each virtual machine in a differential manner (step 303). For example, the resource allocation method can control to distribute the CPU resource and the block I / O resource to each of the virtual machines based on the calculated SLA satisfaction.

In order to control the resource allocation method according to an exemplary embodiment to distribute the calculated SLA satisfaction to the CPU scheduler or the I / O scheduler to control the differential allocation, control is performed so that the CPU resource and the block I / O resources are differentially distributed to the respective virtual machines can do.

Also, in order to control the resource allocation method to control the distribution to be different, the CPU scheduler can control the function of the resource differential distribution by utilizing the function provided by the Linux Complete Fair Scheduler (CFS).

In order to control the allocation of resources to be differentiated, the virtual machines operating in the physical machine are divided into a plurality of groups, and the divided groups are classified into SLA unsatisfied group, SLA supersaturation group, and SLA satisfaction group based on the error value in SLA satisfaction . In addition, the value of quota for each classified group may be calculated, and the resource differentiation may be performed in consideration of the calculated quota value of each group.

In the step of performing the resource differential distribution, a quota value required for the SLAs satisfying the SLAs of the virtual machines in the SLA unsatisfied group to reach 100% may be calculated through the feedback control. In addition, a quota value that the virtual machines belonging to the SLA supersaturation group can transfer to the virtual machine of the SLA dissatisfied group is calculated, and a quota which each virtual machine of the SLA supersaturation group needs to reduce to reach 100% Can be calculated proportionally. Also, a value calculated for the SLA unsatisfied group and each of the virtual machines of the supersaturated group is added to each other by a group, and the two values are compared. A smaller value among the comparison results is selected as a value for quota exchange between the unsatisfied group and the supersaturated group, The value of the quota may be distributed according to the SLA satisfaction rate of the virtual machines belonging to the group.

4 is a diagram showing SLA satisfaction with respect to CPU SLA and I / O SLA by time.

Referring to the graph 400 of FIG. 4, the SLA satisfaction converges from 100% to 100% over time, and the virtual machine 2 (VM2) by the I / O SLA converges to 100% The SLA satisfaction converges from 100% to 100% over time. The resource allocation system according to the present invention can control to distribute CPU resources and block I / O resources to each virtual machine based on the SLA satisfaction calculated so as to be similar to the pattern shown in the graph 400 of FIG. 4 .

FIG. 5 is a diagram showing SLA satisfaction when the GoS framework is used. FIG.

First, the virtual machine 1 (VM1) has a random write 4K Job Type of the CPU and has an SLA Goal of 100 MB / s. Virtual Machine 2 (VM2) has a Random Read 4K Job Type and has an SLA Goal of 200MB / s. The virtual machine 3 (VM3) has the Job Type of the CPU and has SLA Goal of Guarantee 2 cores.

5, the SLA satisfaction of the virtual machine 3 (VM3) exceeds 400% as a pattern when the GoS framework is not used, whereas the virtual machine 1 (VM1) and the virtual machine 2 (VM2) SLA satisfaction can be maintained at around 100%.

5, the SLA satisfaction of the virtual machine 3 (VM3), the SLA satisfaction of the virtual machine 1 (VM1), and the SLA satisfaction of the virtual machine 3 (VM2) SLA satisfaction can be maintained to converge to around 100%.

As a result, by using the present invention, in consideration of CPU and block I / O resources in a KVM-based virtualization environment, each performance requirement of a virtual machine performing a CPU operation and a virtual machine performing a block I / O operation is guaranteed can do.

In addition, the performance requirements of the virtual machine can be achieved by setting the block I / O of the virtual machine and the performance requirements of the CPU to the hypervisor, and distributing the CPU and the block I / O resources with the same reference.

In addition, by using the present invention, since the CPU and the block I / O resources are differentially distributed to the virtual machine through the feedback control, the required performance of the virtual machine can be satisfied for various requirements and situations.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented within a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) unit, a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments and equivalents to the claims are within the scope of the following claims.

100, 200: resource allocation system
110, 210: Statistics Collector
120, 220: SLA manager
130: Differential distribution processor
230: CPU scheduler
240: I / O scheduler

Claims (18)

A statistics collector for collecting performance information of the virtual machine including input / output operations per second (IOPS), I / O bandwidth, latency and CPU utilization;
An SLA manager for calculating SLA (Service Level Agreement) satisfaction required by each virtual machine based on performance information of the collected virtual machines; And
And a differential distribution processor for controlling to distribute the CPU resources and the block I / O resources to the respective virtual machines based on the calculated SLA satisfaction,
The differential distribution processor transfers the computed SLA satisfaction to a CPU scheduler or an I / O scheduler to control the virtual machines to distribute CPU resources and block I / O resources differentially,
Wherein the CPU scheduler performs a resource differential distribution function,
Dividing the virtual machines operating in the physical machine into a plurality of groups and classifying the divided groups into an SLA unsatisfied group, an SLA supersaturation group, and an SLA satisfaction group based on an error value in the SLA satisfaction, And performing a resource differential distribution in consideration of the calculated values of the quota for each group,
Resource allocation system.
The method according to claim 1,
The SLA manager manages SLA value, SLA satisfaction, resource utilization rate used by the virtual machine, and performance information of the virtual machine.
Resource allocation system.
The method according to claim 1,
Wherein the SLA manager provides an API for determining an SLA value and an SLA satisfaction,
Resource allocation system.
delete The method according to claim 1,
The CPU scheduler performs resource differential distribution by utilizing functions provided by CFS (Complete Fair Scheduler) of Linux,
Resource allocation system.
delete The method according to claim 1,
Wherein the CPU scheduler calculates a quota value required for reaching 100% of SLA satisfaction of virtual machines in the SLA unsatisfied group through feedback control,
Resource allocation system.
The method according to claim 1,
Wherein the CPU scheduler calculates a quota value that the virtual machines belonging to the SLA superset group can assign to the virtual machine of the SLA dissatisfied group,
Resource allocation system.
9. The method of claim 8,
Wherein the CPU scheduler calculates a quota value of each virtual machine of the SLA superset group to reduce to reach 100% of the current SLA satisfaction,
Resource allocation system.
9. The method of claim 8,
The CPU scheduler compares values calculated for the SLA dissatisfied group and each of the virtual machines of the supersaturation group by groups and compares the two values, and a smaller value among the comparison results is a value for quota exchange between the unsatisfied group and the supersetting group Choosing,
Resource allocation system.
11. The method of claim 10,
Wherein the CPU scheduler distributes the value of Quota in the SLA unsatisfied group according to the SLA satisfaction rate of the virtual machines belonging to the group and decreases according to the SLA satisfaction rate in the supersaturation group,
Resource allocation system.
The method according to claim 1,
Wherein the I / O scheduler modifies an opportunistic I / O scheduler (OIOS) scheme to feedback control according to SLA satisfaction,
Resource allocation system.
13. The method of claim 12,
Wherein the I / O scheduler performs feedback control of a DAT (Differentially Allowed Time) value indicating a time when a virtual machine can deliver an I / O request to a block I / O device to satisfy a required performance of the virtual machine,
Resource allocation system.
Collecting performance information of a virtual machine, including input / output operations per second (IOPS), I / O bandwidth, latency and CPU utilization, in a statistics collector;
Calculating SLA (Service Level Agreement) satisfaction required by each virtual machine based on performance information of the collected virtual machines in an SLA manager; And
And controlling the differential distribution processor to distribute the CPU resources and the block I / O resources differentially to the respective virtual machines based on the calculated SLA satisfaction,
Wherein the step of controlling to perform the differential distribution includes transferring the calculated SLA satisfaction to a CPU scheduler or an I / O scheduler, and controlling to distribute the CPU resource and block I / O resources to each of the virtual machines,
The step of controlling the differential distribution includes:
Dividing virtual machines operating in a physical machine into a plurality of groups;
Classifying the divided groups into an SLA unsatisfied group, an SLA supersaturation group, and an SLA satisfaction group based on an error value in SLA satisfaction; And
Calculating a quota value of each classified group and performing a differential distribution of resources by taking into account the calculated value of the quota for each group;
Resource allocation method.
delete 15. The method of claim 14,
Wherein the step of controlling the differential distribution includes controlling the CPU scheduler to perform a resource differential distribution function by utilizing a function provided by the Complete Fair Scheduler (CFS) of Linux.
Resource allocation method.
delete 15. The method of claim 14,
Wherein performing the differential distribution of resources comprises:
Calculating a quota value required for the SLA satisfaction of the virtual machines in the SLA unsatisfied group to reach 100% through feedback control;
A quota value that the virtual machines belonging to the SLA oversaturated group can transfer to the virtual machine of the SLA unsatisfied group is calculated and the quota value of each virtual machine in the SLA superset group is reduced to reach 100% The value is calculated proportionally,
A value calculated for each SLA unsatisfied group and each of the virtual machines in the supersaturated group is added to each group to compare two values, and a smaller value among the comparison results is selected as a value for quota exchange between the unsatisfied group and the supersaturated group,
And distributing the value of the selected Quota according to the SLA satisfaction rate of the virtual machines belonging to the group.
Resource allocation method.

Images