KR101852608B1 - Virtual machine monitor, schduling method of virtual machine monitor and computer readable medium of thereof - Google Patents

Abstract

Disclosed are a virtual machine monitor, a virtual machine monitor scheduling method, and a computer-readable recording medium including computer program therefor. The virtual machine monitor includes: a scheduler operating at least two domains; a task characteristic evaluating part; and a credit management part. According to the method of scheduling the virtual machine monitor, the scheduler schedules the domains and executes a boost mechanism for the domains when an even of the domains is delayed, and once the boost mechanism is executed, the task characteristic evaluating part calculates an I/O distribution ratio of the domains and an average boost use time, which is an average CPU use time for the execution of the boost mechanism, and then evaluates task characteristics of the domains based on the calculated information, and the credit management part determines a credit allocation amount and a repayment period in accordance with the evaluated task characteristics. According to the present invention, when the boost mechanism is executed, consumption patterns of a boost credit and an under credit are analyzed in relation to the mechanism, and then, task characteristics are evaluated based on the patterns and a credit allocation amount and repayment period are determined in accordance with the evaluated characteristics, and thus, a differential and dynamic boost mechanism is able to be applied. As such, the present invention is capable of improving I/O responsibility and fairly distributing CPU resources.

Description

TECHNICAL FIELD [0001] The present invention relates to a scheduling method for a virtual machine monitor, a virtual machine monitor, and a computer-readable recording medium on which a computer program for providing the same is recorded.

The present invention relates to a scheduling method of a virtual machine monitor, a virtual machine monitor, and a computer-readable recording medium on which a computer program for providing the same is recorded. More specifically, the present invention relates to a scheduling method for improving the I / O responsiveness of scheduling of a virtual machine A virtual machine monitor, a scheduling method of a virtual machine monitor, and a computer-readable recording medium on which a computer program for providing the same is recorded.

Virtualization refers to the efficient use of resources by consolidating multiple physical resources into logical resources or conversely by dividing a single physical resource into multiple logical resources. Virtualization technology can introduce a virtualization layer to run multiple systems on a single machine. Software that abstracts physical resources is called a virtual machine monitor or a hypervisor. Some of the virtual machine monitors that we often encounter are Hyper-V, VirtualBox, VMware, Xen, and KVM.

A virtual machine is a logical resource that is abstracted by a virtual machine monitor, that is, a machine environment, such as a physical machine that is not a physical machine. Xen also refers to a virtual machine as a domain. The relationship between a virtual machine monitor and a virtual machine is similar to that of an operating system and a process. The operating system abstracts physical resources and provides them to multiple processes. Similarly, virtual machine monitors also virtualize CPU, memory, and I / O to provide virtual machines.

VCPU refers to a logical abstraction of a CPU, which is a real physical resource, and a plurality of virtual machines running on a single physical machine have a VCPU, not a real CPU. The VCPU corresponds to the process in the operating system-process relationship. As the operating system selects the process to use the CPU resource, the virtual machine monitor also selects the VCPU to use the CPU resource through the scheduling policy. The scheduling that determines the VCPU to use CPU resources is called virtual machine scheduling.

On the other hand, in order to process I / O events in the credit scheduler, the corresponding VCPUs must be scheduled in the waiting queue. At this time, there is no task in the domain, and the VCPU having 300 credits or more is idled or only I / O intensive work is performed, and the VCPU becomes inactive. In this case, the VCPUs are deactivated because they are excluded from the waiting queue of the credit scheduler, and thus a delay occurs in the I / O event responsiveness until they are activated.

The boost mechanism is a way to resolve the delay for I / O event responsiveness for this deactivated VCPU. When the event is delayed to the VCPU in the inactive state, the boost mechanism changes the corresponding VCPU to the executable state and gives the highest priority BOOST. The VCPU of the boost priority order is executed first, and the delayed event is rapidly processed to improve the responsiveness of the domain.

However, such a boost mechanism has a problem in that the inactive VCPU has the highest boost priority and is the first to be scheduled, rather, the I / O event responsiveness of the activated VCPU is lowered. In addition, when there are many CPU-intensive active VCPUs in the scheduler's wait queue, the I / O event responsiveness of other activated VCPUs is lowered.

Korea Patent Publication No. 2011-0103257 Korea Patent Publication No. 2012-0035387

According to an aspect of the present invention, when the boost mechanism is executed for a domain, an I / O distribution ratio and a boost average use time are calculated to evaluate a task characteristic of a domain, and a credit allocation amount and a repayment period are determined according to an evaluated task characteristic And a scheduling method of the virtual machine monitor and the virtual machine monitor.

A scheduling method of a virtual machine monitor according to an aspect of the present invention is a scheduling method of a virtual machine monitor including a scheduler that drives at least two domains, a task characteristic evaluation unit, and a credit management unit. The scheduler schedules the domain, When the event of the domain is delayed, executes a boost mechanism for the domain, and when the boost mechanism is executed, the task characteristic evaluator calculates a boost ratio of the domain, which is the CPU average time used in the execution of the boost mechanism, Calculates the average usage time, evaluates the task characteristics of the domain based on the average usage time, and the credit management unit determines the credit allocation amount and the repayment period according to the evaluated task characteristic.

The method may further include classifying the credit when the scheduler executes the boost mechanism, wherein the credit includes a credit variable defining a credit allocation of the first cycle from the start of credit consumption until the next boost mechanism is executed, A boost credit variable that defines a boost credit quota that can be used for the boost mechanism among the quota quantities and an undercredit variable that defines an amount of under credit that can be used in addition to the boost mechanism among the credit quota.

Also, the credit variable is defined by Equation (1) below, and the boost credit variable is defined by Equation (2) below, and the under credit variable can be defined by Equation (3) below.

[Equation 1]

Here, c _{i, j} (t) is a credit variable, w _i is a weight value applied at the time of allocating additional credit, ACCUM is a period in which additional credit is allocated, t _{i, j,} T _{i, j, s} is a time point at which consumption of at least one of the boost credit variable and the under credit variable is completed, and t _{i, j, r} is a time when the reimbursement for the consumed boost credit is completed T _{i, j +1,1} is the start of the second period, and nonboost Time _i is the CPU time used without applying the boost mechanism during the first period.

&Quot; (2) "

_{J i} (t _{i, j, 1} ) is the amount of credit allocated at the beginning of the first period, and io _{i, j} is the amount of the task characteristic I / O distribution ratio calculated by the evaluation unit, and boostTime _i is the CPU time used by the boost mechanism during the first period.

&Quot; (3) "

Where u _{i, j} (t) is the under-credit variable.

In addition, the task characteristic evaluation unit calculates the I / O distribution ratio of the domain,

Calculating the consumption amount of the boost credit in the first period and the consumption amount of the under credit and calculating an I / O distribution ratio of the second period as a ratio of the consumption amount of the boost credit to the total credit consumption.

Further, the I / O distribution ratio can be calculated by the following equation.

Here, io _{i, j} denotes an I / O division ratio, and when j = 1, the first time the boost mechanism is applied, and? B _{i, j - 1} and? U _{i, j -1} denote the boost The credit consumption amount and the under-credit consumption amount.

In addition, the task characteristic evaluation section calculates the boost average use time,

Based on the CPU time applied by the boost mechanism to the number of times the boost mechanism during the first period is applied.

Also, the credit managing unit determines the credit allocation amount and the redeeming period according to the evaluated task characteristic, based on the I / O distribution ratio and the boost average usage time, the redeemable rate used for the redemption, And < / RTI >

Here, rr _{i, j} is the repayment rate, io _{i, j} is the I / O distribution ratio, and avgboostTime _i means the boost average usage time.

Further, a computer-readable recording medium on which a computer program is recorded for providing a virtual machine monitor according to the above-described method may be included.

A virtual machine monitor according to another aspect of the present invention is a virtual machine monitor that drives at least two or more domains and includes a scheduler for scheduling the domain and executing a boost mechanism for the domain when an event of the scheduled domain is delayed, A task characteristic evaluation unit for calculating an I / O distribution ratio of the domain and a CPU average time used in the boost mechanism when the mechanism is executed, and evaluating a task characteristic of the domain based on the I / O distribution ratio, And determines a credit allocation amount and a redeeming time based on the credit allocation.

Meanwhile, the scheduler classifies the credit when the boost mechanism is executed, wherein the credit is a credit variable that defines a credit allocation amount of the first cycle from the start of credit consumption until the next boost mechanism is executed, A boost credit variable that defines a boost credit quota that can be used in the boost mechanism and an undercredit variable that defines an amount of under credit that can be used in addition to the boost mechanism among the credit quota.

Also, the credit variable is defined by Equation (1) below, and the boost credit variable is defined by Equation (2) below, and the under credit variable can be defined by Equation (3) below.

[Equation 1]

Here, c _{i, j} (t) is a credit variable, w _i is a weight value applied at the time of allocating additional credit, ACCUM is a period in which additional credit is allocated, t _{i, j,} T _{i, j, s} is a time point at which consumption of at least one of the boost credit variable and the under credit variable is completed, and t _{i, j, r} is a time when the reimbursement for the consumed boost credit is completed T _{i, j +1,1} is the start of the second period, and nonboost Time _i is the CPU time used without applying the boost mechanism during the first period.

&Quot; (2) "

_{J i} (t _{i, j, 1} ) is the amount of credit allocated at the beginning of the first period, and io _{i, j} is the amount of the task characteristic I / O distribution ratio calculated by the evaluation unit, and boostTime _i is the CPU time used by the boost mechanism during the first period.

&Quot; (3) "

Where u _{i, j} (t) is the under-credit variable.

Also, the task characteristic evaluation unit may calculate the consumption amount of the boost credit and the consumption amount of the under credit in the first cycle, and determine the I / O distribution ratio of the second cycle as the ratio of the consumption amount of the boost credit to the total credit consumption .

Further, the I / O distribution ratio can be defined by the following equation.

Here, io _{i, j} denotes an I / O division ratio, and when j = 1, the first time the boost mechanism is applied, and? B _{i, j - 1} and? U _{i, j -1} denote the boost The credit consumption amount and the under-credit consumption amount.

Also, the task characteristic evaluation unit may calculate the CPU time used by applying the boost mechanism to the number of times the boost mechanism during the first period is applied, and calculate the boost average use time.

Also, the credit managing unit may determine a repayment ratio to be used for repayment of the additional credit based on the boost average usage time as shown in the following equation.

Here, rr _{i, j} is the repayment rate, io _{i, j} is the I / O distribution ratio, and avgboostTime _i means the boost average usage time.

According to the present invention, when the boost mechanism is executed, the consumption patterns of the boost credit and the under credit are analyzed, the task characteristic is evaluated based on the consumption pattern, and the credit allocation amount and the repayment period are determined according to the evaluated task characteristic. The dynamic boost mechanism can be applied. This can ultimately improve I / O responsiveness and enable fair distribution of CPU resources.

1 is a diagram illustrating a separate driver model including a virtual machine monitor according to an embodiment of the present invention.
2 is a diagram showing a specific configuration of the virtual machine monitor shown in FIG.
3 is a diagram showing a credit lifecycle executed by the scheduler.
4 is a diagram showing task characteristics of the domain evaluated by the task characteristic evaluation unit.
5 is a flowchart illustrating a scheduling method of a virtual machine monitor according to another embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The advantages and features of the present invention, and the manner of achieving them, will be apparent from and elucidated with reference to the embodiments described hereinafter in conjunction with the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Is provided to fully convey the scope of the invention to those skilled in the art, and the invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

The terminology used herein is for the purpose of illustrating embodiments and is not intended to be limiting of the present invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. It is noted that the terms " comprises "and / or" comprising ", as used herein, do not exclude the presence or addition of one or more other elements, steps and operations.

1 shows a separate driver model including a virtual machine monitor according to an embodiment of the present invention.

Referring to FIG. 1, a separate driver model may include a plurality of domains 100, 200, a virtual machine monitor 300, and a hardware platform 400.

The domains 100 and 200 may be implemented as an operating system or an application program so as to provide an environment that is not a real machine but a physical machine. These domains 100 and 200 can be divided into a driver domain 100 and a guest domain 200. The driver domain 100 includes domains 100 and 200 that can directly access the hardware platform 400 and exchange I / And the guest domain 200 refers to the domains 100 and 200 that indirectly use the hardware platform with the help of the driver domain 100. [

The hardware platform 400 may be a physical device such as a central processing unit (CPU), a memory, an input / output port, etc. and may exchange direct I / O event information with the driver domain 100.

The virtual machine monitor 300 is a platform for simultaneously operating a plurality of domains 100 and 200 in a single machine, and can manage abstraction of hardware resources or distribute resources to domains through a virtual machine scheduler. That is, the virtual machine monitor 300 is software that enables a plurality of operating systems to be simultaneously operated on one computer.

The virtual machine monitor 300 can use various virtual machine schedulers, and is typically a real time deferrable server (RTDS), ARINC 653, credit, and credit 2 scheduler provided by the Xen Project VMM. Among these, the credit scheduler is a general purpose scheduler, which is a proportional process equity scheduler that distributes CPU resources in proportion to the equity of the domains (100, 200).

The credit scheduler distributes the credit to the VCPUs of the domains 100 and 200 in proportion to the weight value of the domains 100 and 200 every 30 ms. The VCPU is prioritized according to the value of the credit held, and the credit is deducted as much as the executed credit .

The credit scheduler performs a boost mechanism that assigns a VCPU with a delayed I / O event to the BOOST state, which is the highest priority, when the I / O event is delayed in the VCPU. In this case, the boost mechanism sets the VCPU of the inactive domain (100,200) Since the BOOST state is given to the VCPU only when the / O event is delayed, the VCPUs of the active domains 100 and 200 are hardly given the BOOST state. In addition, when there are many CPU-intensive active VCPUs in the scheduler's wait queue, the I / O event responsiveness of other activated VCPUs is lowered.

Accordingly, the present invention provides a differential and dynamic new boost mechanism by analyzing the credit consumption pattern when the boost mechanism is executed, evaluating the task characteristic, and determining the credit and the repayment period to be allocated in the next boost mechanism according to the estimated task characteristic .

To this end, the present invention classifies the credit as a credit variable, a boost credit variable and an under credit variable in order to analyze the credit consumption pattern according to the credit life cycle. The analysis of the credit consumption pattern may include determining an I / O distribution ratio to be applied to the second boost mechanism based on the boost credit variable and the under-credit variable consumed in the first boost mechanism, and comparing the number of times the boost mechanism is applied And calculates the boost average usage time based on the CPU time that has been consumed. Evaluates the task characteristic based on the analyzed credit consumption pattern, and determines the credit allocation amount and the redeeming time in the next cycle in which the second boost mechanism starts according to the task characteristic divided into the predetermined criteria.

To this end, the virtual machine monitor 300 of the present invention executes the credit / repayment mechanism of the credit, calculates the I / O distribution ratio and the boost average usage time of the domains 100 and 200 using a loan / redemption mechanism, Evaluates the task, and determines a reimbursement rate to be used for reimbursement among the additional credits based on the evaluated task characteristic, and executes a new boost mechanism in which the credit quota and reimbursement period are adjusted.

The virtual machine monitor 300 will be described in detail with reference to FIG.

Referring to FIG. 2, the virtual machine monitor 300 includes a scheduler 310, a task characteristic evaluation unit 320, and a credit management unit 330.

More specifically, the scheduler 310 schedules a domain and performs a boost mechanism for the domain if an event of the scheduled domain is delayed. To this end, the scheduler 310 schedules based on a lending / redemption mechanism. The lending / repayment mechanism will be described as follows. First, when the I / O event is delayed in the VCPU of the domain, the scheduler lends the loan credit to the VCPU and applies the boost mechanism. The loan credit is deducted as much as the VCPU is boosted and executed, and the scheduler redeems the loan credit used in the later added credit and allocates the remaining credit.

Particularly, the scheduler 310 according to the present embodiment classifies the credit after the boost mechanism is executed into three kinds of credit variables according to the loan / redemption mechanism described above, and schedules the credit lifecycle as shown in FIG.

3, the first, the scheduler 310 first VCPU from the time (t _{i, j, 1)} who applies the boost mechanism the next boost mechanism is executed (t _{i, j +1, 1)} prior to the time that A credit variable that defines the credit quota of the first cycle (t _{i, j, 1} to t _{i, j +1, 1} ) in the first cycle and a boost credit that defines the credit quota used in the boost mechanism Variable and an under-credit variable that can be used in addition to the boost mechanism among the above credit allocations, that is, an under-credit quota for CPU operation processing.

Once the credit quota is defined, credit consumption begins at the time the VCPU first applies the boost mechanism (t _{i, j, 1} ). When the credit consumption of at least one of the boost credit variable and the under credit variable is completed after the credit consumption is started, up to the time t _{i, j, s} is defined as the consumption period sp _{i, j} . And have paid off a by the consumption boost credit amount disclosed for since the consumption period (sp _{i, j),} the consumption period, the redemption start (t _{i, j, s)} completing repayment from (t _{i, j, r)} when (Rp _{i, j} ) are defined.

The definition of each credit variable according to the above-mentioned credit life cycle is as follows.

First, a credit variable defining a credit allocation amount up to the first period (t _{i, j, 1} to t _{i, j +1, 1} ) is defined according to Equation (1).

[Equation 1]

Here, c _i, and _j (t) is a credit variable and, w _i is the weight value to be applied when additional credit allocation, ACCUM is (can be, for example, 30ms) periods taking allocation of additional _credit, t _{i, j, 1} is the start of the first period, t _{i, j, s} is the time when consumption of at least one of the boost credit variable and the under credit variable is completed, t _{i, j,} The nonboost Time _i defines the CPU time used without applying the boost mechanism during the first period.

That is, the credit parameter is the first point of time when receiving the VCPU applying the boost mechanism (t _{i, j, 1)} from the next boost mechanism is executed (t _{i, j +1, 1)} to the first cycle (t _{i, j} up to the previous _{, 1} to t _{i, j + 1,1} ), respectively.

First, during the consumption period (sp _{i, j} : t _{i, j, 1} to t _{i, j, s} ) _, additional credit is assigned to each ACCUM as much as the weight value once.

2) During the redemption period (rp _{i, j} : t _{i, j, s} to t _{i, j, r} ) _, additional credits are allocated for each ACCUM No additional CPU time (nonboostTime _i ) except for the additional credit is allocated. Here, the repayment rate is a rate used for repayment among additional credits, which is calculated by the credit management unit, and the calculation method is explained later.

3) Additional credits are assigned to ACCUMs as much as the weight value once before the recharging is completed and before the next boost mechanism is executed (t _{i, j, r} to t _{i, j + 1,1} ) No more CPU time (nonboostTime _i ), but additional credit

Also, the boost credit variable that defines the boost credit that can be used for the boost mechanism among the credit allocations up to the first period (t _{i, j, 1} to t _{i, j +1, 1} ) do.

&Quot; (2) "

_Here, b _{i, j} (t) is the boost credit variables, io _{i, j} is the I / O allocation ratio, boostTime _i is the first cycle _{(t i, j, 1 ~} t i, j + 1, ₁ ) < / RTI > the CPU time that the boost mechanism is applied to.

(T _{i, j, 1} ) at the start of the consumption period (sp _{i, j} : t _{i, j, 1} to t _{i, j, s} ) The initial boost credit is allocated as much as the quota is proportional to the I / O distribution ratio. Here, the I / O distribution ratio is calculated from the consumption pattern of the previous consumption period as a ratio of the boost credit to be distributed to the consumption time. The method of calculating the I / O distribution ratio will be described later. During the subsequent consumption period (sp _{i, j} : t _{i, j, 1} to t _{i, j, s} ), the boost mechanism in the initially allocated boost credit is applied and subtracted by the boost time (boostTime) Is defined. 3) Also, during the repayment period (rp _{i, j} : t _{i, j, s} to t _{i, j, r} ), the consumption period is completed and is defined as '0'.

On the other hand, the first cycle (t _{i, j, 1} ~ t _{i, j +1, 1)} Credits allocation of under-credit variables that define the under credit quota that may be used in addition to the boost mechanism is according to the formula 3; Is defined.

&Quot; (3) "

That is, with respect to the under credit variables: 1) consumption period (sp _{i, j:} t _{i, j, 1} ~ t _{i, j, s),} time (t _{i, j, 1)} at the start, the credit of the time The initial undercredit is allocated as much as the quota credit except for the boost credit, and 2) during the subsequent consumption period (sp _{i, j} : t _{i, j, 1} to t _{i, j, s} ) (NonboostTime _i ) which is not used and under-priority. 3) Also, during the redemption period, the consumption period is completed and is defined as '0'.

Scheduler 310 includes a first consumption cycle period _{(t i, j, 1 ~} t i, j +1, 1) (sp i, j: t i, j, 1 ~ t i, j, s) is complete The boost time, the boost time, the CPU time used by the boost mechanism, and the boost mechanism during the consumption period (sp _{i, j} : t _{i, j, 1} to t _{i, j, s} ) And transmits the boost count information, which is the received count information, to the task characteristic evaluation unit 320. [

Based on the information received from the scheduler 310, the task characteristic evaluation unit 320 calculates the I / O distribution ratio of the domain and the average CPU utilization time, which is the CPU average time used in the boost mechanism, .

Specifically, the task characteristics evaluating unit 320 includes a first consumption cycle period (t _{i, j, 1} ~ t _{i, j +1, 1)} (sp _{i, j:} t _{i, j, 1} ~ t _i, the boost credits (Δb _{i, j)} and the under-credit amount (Δu _{i, j)} used for a _{j, s)} is calculated by the equation (4) below.

&Quot; (4) "

Also, the task characteristic evaluation unit 320 calculates the I / O distribution ratio of the second period by the ratio of the total credit consumption amount and the boost credit consumption amount, and the calculated I / O distribution ratio is defined by Equation (5).

&Quot; (5) "

Here, io _{i, j} denotes an I / O division ratio, and when j = 1, the first time the boost mechanism is applied, and? B _{i, j - 1} and? U _{i, j -1} denote the boost The credit consumption amount and the under-credit consumption amount.

That is, when the first boost mechanism is applied, consumption patterns in the previous consumption period are not known, so boost credit and under credit are allocated in half of the initial allocated credit amount. In the subsequent cycle, VCPUs that perform I / O-oriented operations reflecting the consumption patterns in the previous consumption period receive more boost credit ratios and VCPUs that perform CPU-based tasks are assigned more under credit ratios have. Thus, task characteristics can be tracked dynamically and credit resources can be efficiently distributed to the characteristics of the VCPU task.

The task characteristic evaluation unit 320 calculates a boost average usage time based on the CPU time applied to the number of times that the boost mechanism during the first period is applied to the boost mechanism. The calculation method of the boost average use time is given by Equation (6) below.

&Quot; (6) "

Where avgboostTime is the average time to use the boost, boostTime is the CPU time used by the boost mechanism from t1 to t2, and boostCount is the number of times the boost mechanism from t1 to t2 has been applied. On the other hand, the boost average usage time means an average usage time during the application of the boost mechanism, so t1 to t2 may mean the consumption period during the first period.

The task characteristic evaluation unit 320 classifies the task characteristics as shown in FIG. 4 on the basis of the I / O distribution ratio and the boost average use time, and classifies the tasks classified based on the calculated I / O distribution ratio and the boost average usage time It is possible to evaluate which one of the characteristics belongs.

First, the task characteristic will be described. Task characteristics can be classified as I / O intensive or CPU intensive. The higher the I / O concentration, the higher the boost credit ratio compared to the under credit, and the higher the CPU concentration, the higher the under credit ratio compared to the boost credit. Therefore, although the VCPU having the task of the I / O-centric task can receive the boost mechanism more, the VCPU having the task with the high CPU concentration may not be scheduled due to the priority being pushed. In the case of a VCPU having an I / O and a CPU-intensive hybrid task, CPU resources may be monopolized in scheduling. During the repayment period, the VCPUs having tasks with high CPU concentration can be scheduled because the boost mechanism has not been applied and the priorities have been pushed out so that the credit management unit 330 can use the additional credit allocation amount And determines the credit allocation amount and the repayment time.

The characteristics of the task can be classified according to whether they are CPU intensive or I / O intensive. The task characteristics can be classified as shown in FIG. 4 according to the I / O distribution ratio and the average time of use, which is the average time used by the boost mechanism during the consumption period. Referring to FIG. 4, the task characteristic can be classified into A, B, C, and D types by grasping the degree of task concentration based on the I / O distribution ratio and the boost average usage time. In the classification graph of FIG. 4, the X axis is the boost average use time and the Y axis is the I / O distribution ratio. The high I / O distribution ratio means that the VCPUs in the current consumption period have a lot of boost credits consumed by the boost mechanism, which means that the I / O concentration is high. In addition, the fact that the average CPU usage time is high means that the CPU operation is performed in addition to the I / O operation, which means that the CPU concentration is high. Therefore, in the graph of FIG. 4, the A type indicates a task having a high I / O concentration, and the B type indicates a composite task having a high I / O concentration and a CPU concentration. Also, C type means idle task with low I / O concentration and CPU concentration, and D type means CPU intensive task.

The credit managing unit 330 determines the credit allocation amount and the redeeming time based on the task characteristic evaluated by the task characteristic evaluating unit 320. [ Specifically, the credit management unit 330 determines the time required to repay the boost credit, that is, the repayment rate, in accordance with the above-described four types.

The credit managing unit 330 determines the repayment ratio, which is the ratio of the credit used for the redemption among the additional credits according to Equation (7) below.

&Quot; (7) "

Here, rr _{i, j} is the repayment rate, io _{i, j} is the I / O distribution ratio, and avgboostTime _i means the boost average usage time calculated by Equation (6). That is, the credit managing unit 330 determines that the repayment rate is proportional to the I / O distribution ratio and inversely proportional to the boost average usage time.

A process for determining the credit allocation amount and the repayment period of the credit management unit 330 together with the task type of FIG. 4 will be described below.

First, since the I / O intensive A and B types have consumed a lot of boost credits, the credit managing unit 330 determines a high amount of credit required for redemption. On the other hand, type B is not only I / O intensive, but also CPU-intensive, so that more CPU resources are monopolized when the boost mechanism is applied than type A. Accordingly, the credit managing unit 330 determines the repayment rate so that the B type redeeming time is relatively longer than that of the A type, and the CPU resources are distributed equitably. In the case of the C type, the credit managing unit 330 regards the I / O concentration and the CPU concentration as low, and therefore, it is regarded as an idle task. Also, in the case of the CPU-intensive D type, the credit managing unit 330 allocates the boost credit to a small amount and makes the repayment time longer, so that the number of boost mechanisms is less than that of the A and B types.

5 is a flowchart illustrating a scheduling method of a virtual machine monitor according to another embodiment of the present invention.

Referring to FIG. 5, a scheduling method of a virtual machine monitor according to an embodiment of the present invention is a scheduling method in a virtual machine monitor including a scheduler for driving at least two domains, a task characteristic evaluation unit, and a credit management unit, When an event of the scheduled and scheduled domain is delayed 510, a boost mechanism for the domain is applied (520).

When the boost mechanism is applied (520), the task characteristic evaluation unit calculates the I / O distribution ratio as the result of applying the boost mechanism, and calculates the boost average usage time, which is the CPU average time used in the execution of the boost mechanism (530) (540). ≪ / RTI >

The credit management unit determines the credit allocation amount and the repayment period in the next cycle according to the evaluated task characteristic (550).

According to the present invention, when the boost mechanism is executed, the consumption patterns of the boost credit and the under credit are analyzed, the task characteristic is evaluated based on the consumption pattern, and the credit allocation amount and the repayment period are determined according to the evaluated task characteristic. Dynamic bushing mechanism can be applied. This can ultimately improve I / O responsiveness and enable fair distribution of CPU resources.

As such, the virtual machine monitor 300 and virtual machine monitor 300 scheduling methods may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program commands, data files, data structures, and the like, alone or in combination.

The program instructions recorded on the computer-readable recording medium may be ones that are specially designed and configured for the present invention and are known and 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 recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, 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 generated 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 device may be configured to operate as one or more software modules for performing the processing according to the present invention, and vice versa.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, You will understand. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive.

300: Virtual Machine Monitor
310: Scheduler
320: task characteristic evaluation unit
330: Credit management part

Claims (15)

A scheduling method of a virtual machine monitor including a scheduler for driving at least two domains, a task characteristic evaluation unit, and a credit management unit,
Wherein the scheduler schedules the domain and performs a boost mechanism for the domain when an event of the scheduled domain is delayed,
Wherein when the boost mechanism is executed, the task characteristic evaluation unit calculates an I / O distribution ratio of the domain and a boost average use time which is a CPU average time used when the boost mechanism is executed, evaluates a task characteristic of the domain based on the calculated average,
Wherein the credit managing unit determines a credit allocation amount and a redeeming period according to the evaluated task characteristic,
The scheduler classifies the credit when the boost mechanism is executed,
The credit may include:
A credit variable for defining a credit allocation of the first cycle from the start of the credit consumption until the next boost mechanism is executed,
A boost credit variable that defines a boost credit quota that can be used for the boost mechanism among the credit quota,
And an under-credit variable that defines an amount of under-credit that can be used in addition to the boost mechanism,
Wherein when the boost mechanism is applied for the first time, the boost credit and the under credit are allocated in half the amount of the initially allocated credit. delete The method according to claim 1,
Wherein the credit variable is defined by Equation (1) below, the boost credit variable is defined by Equation (2) below, and the under-credit variable is defined by Equation (3) below.
[Equation 1]

Here, c _{i, j} (t) is a credit variable, w _i is a weight value applied at the time of allocating additional credit, ACCUM is a period in which additional credit is allocated, t _{i, j,} T _{i, j, s} is a time point at which consumption of at least one of the boost credit variable and the under credit variable is completed, t _{i, j, r} is a time when the reimbursement for the consumed boost credit is completed T _{i, j + 1,1} is the start of the second period, and nonboost Time _i is the CPU time used without applying the boost mechanism during the first period.
&Quot; (2) "

_{J i} (t _{i, j, 1} ) is the amount of credit allocated at the beginning of the first period, and io _{i, j} is the amount of the task characteristic I / O distribution ratio calculated by the evaluation unit, and boostTime _i is the CPU time used by the boost mechanism during the first period.
&Quot; (3) "

Where u _{i, j} (t) is the under-credit variable. The method of claim 3,
The task feature evaluator calculates the I / O distribution ratio of the domain by:
Calculating a consumption amount of the boost credit and a consumption amount of the under credit in the first cycle,
And calculating an I / O distribution ratio of the second cycle at a ratio of the total credit consumption to the boost credit consumption. 5. The method of claim 4,
Wherein the I / O distribution ratio is calculated according to the following equation.

Here, io _{i, j} denotes an I / O division ratio, and when j = 1, the first time the boost mechanism is applied, and? B _{i, j - 1} and? U _{i, j -1} denote the boost The credit consumption amount and the under-credit consumption amount. The method of claim 3,
The task characteristic evaluating section calculates the boost average use time,
Wherein a boost mechanism is applied to the number of times the boost mechanism during the first period is applied, and is calculated based on the CPU time used. The method according to claim 6,
Wherein the credit managing unit determines the credit allocation amount and the redeeming period according to the evaluated task characteristic,
Based on the I / O distribution ratio and the boost average usage time, determining a redemption rate to be used for redemption among additional credits as shown in the following equation.

Here, rr _{i, j} is the repayment rate, io _{i, j} is the I / O distribution ratio, and avgboostTime _i means the boost average usage time. A computer-readable recording medium having recorded thereon a computer program for causing a computer to execute a scheduling method of a virtual machine monitor according to any one of claims 1 to 9. 1. A virtual machine monitor for driving at least two domains,
A scheduler that schedules the domain and performs a boost mechanism for the domain when an event of the scheduled domain is delayed;
A task characteristic evaluation unit for calculating a boost average usage time which is an average CPU time used when the boost mechanism is executed and an I / O distribution ratio of the domain when the boost mechanism is executed, and evaluating a task characteristic of the domain based on the calculated average boost time; And
And a credit managing unit for determining a credit allocation amount and a redeeming time based on the task characteristic,
The scheduler classifies the credit when the boost mechanism is executed,
The credit may include:
A credit variable for defining a credit allocation of the first cycle from the start of the credit consumption until the next boost mechanism is executed,
A boost credit variable that defines a boost credit quota that can be used for the boost mechanism among the credit quota,
And an under-credit variable that defines an amount of under-credit that can be used in addition to the boost mechanism,
Wherein when the boost mechanism is applied for the first time, the boost credit and the under credit are allocated in half of the initially allocated credit amount. delete [Claim 11 is abandoned upon payment of the registration fee.] 10. The method of claim 9,
Wherein the credit variable is defined by Equation (1) below, the boost credit variable is defined by the following Equation (2), and the under-credit variable is defined by Equation (3) below.
[Equation 1]

Here, c _{i, j} (t) is a credit variable, w _i is a weight value applied at the time of allocating additional credit, ACCUM is a period in which additional credit is allocated, t _{i, j,} T _{i, j, s} is a time point at which consumption of at least one of the boost credit variable and the under credit variable is completed, t _{i, j, r} is a time when the reimbursement for the consumed boost credit is completed T _{i, j + 1,1} is the start of the second period, and nonboost Time _i is the CPU time used without applying the boost mechanism during the first period.
&Quot; (2) "

_{J i} (t _{i, j, 1} ) is the amount of credit allocated at the beginning of the first period, and io _{i, j} is the amount of the task characteristic I / O distribution ratio calculated by the evaluation unit, and boostTime _i is the CPU time used by the boost mechanism during the first period.
&Quot; (3) "

Where u _{i, j} (t) is the under-credit variable. [12] has been abandoned due to the registration fee. 12. The method of claim 11,
The task characteristic evaluation unit
Calculating a consumption amount of the boost credit and a consumption amount of the under credit in the first cycle,
A virtual machine monitor that determines an I / O distribution ratio of a second cycle at a ratio of the total credit consumption to the boost credit consumption. [13] has been abandoned due to the registration fee. 13. The method of claim 12,
Wherein the I / O distribution ratio is defined by the following equation:

Here, io _{i, j} denotes an I / O division ratio, and when j = 1, the first time the boost mechanism is applied, and? B _{i, j - 1} and? U _{i, j -1} denote the boost The credit consumption amount and the under-credit consumption amount. [14] has been abandoned due to the registration fee. 12. The method of claim 11,
The task characteristic evaluation unit
Wherein the controller calculates the CPU time by applying a boost mechanism to the number of times the boost mechanism is applied for the first period, and calculates the boost average usage time. [Claim 15 is abandoned upon payment of registration fee] 15. The method of claim 14,
The credit management unit,
And determines a repayment ratio to be used for repayment of the additional credit based on the boost average usage time as shown in the following equation.

Here, rr _{i, j} is the repayment rate, io _{i, j} is the I / O distribution ratio, and avgboostTime _i means the boost average usage time.

Images