JP6732693B2 - Resource allocation control system, resource allocation control method, and program - Google Patents

Description

The present invention relates to a technique of sharing a plurality of resources provided by a service such as a cloud with a plurality of users.

In recent years, a service model represented by cloud services in which resources are shared by a plurality of users has become common. By sharing computer resources such as CPU, memory, and storage, and network resources such as link bandwidth among multiple users, it is possible to meet demand with a smaller amount of resources compared to the model in which resources are prepared for each user. It is a service model.

However, when resources are shared among users, individual user demands (which may be referred to as requests) may reach resource limits. In preparation for such a case, in the resource sharing model, allocation control is required to determine whether or not to accept (allocate) a resource to the demand of the user or how much resource to allocate to the user. Becomes

Here, the requirements for allocation control include resource utilization efficiency and inter-user fairness. When one user occupies a resource, another user is in demand, and when the user is not assigned to the resource, unfairness occurs. On the other hand, if an upper limit is set on the resources allocated to a user to avoid unfairness, even if there is no demand from other users, the user will not be allocated more resources than the upper limit, and resource utilization efficiency will increase. There is a problem that the sex decreases.

Furthermore, when a plurality of resource types are assigned, it is necessary to assign the resource types in consideration of the demand ratio of the user to the resource types and whether the resource demand is elastic (elastic).

For example, when the resource type is two types, CPU and memory, when a task of one user requires a large amount of CPU and a task of another user requires a large amount of memory, the demand ratio of the user is different, so Evenly dividing resources is not efficient from the viewpoint of user task execution. In view of such a problem, in Non-Patent Document 1, the main resource type (Dominant Resource) is determined for each user from the demand ratio thereof, and the main resource is allocated so as to be fair with respect to the main resource (Dominant Resource Fairness). Techniques are disclosed. Also disclosed is a method called asset fairness, which fairs not only the main resources but also the sum of occupation rates of all resources.

The method in this document assumes an environment in which each user has tasks to be executed indefinitely and uses up all the allocated resources, and is a method of allocating resources so that major resources are fair up to the resource capacity upper limit. is there. However, in a general cloud environment, the resource demand of each user is finite, and there is a possibility that resources will not be used in the allocation by the same method.

Furthermore, the document assumes that resource demand is not elastic. Here, that the resource demand is elastic refers to the characteristic that the demand of the user can be satisfied to some extent even when the resource less than the demand of the user is allocated. For example, when the resource type is CPU, even when n-1 CPUs are allocated when the demand is n CPUs, the task execution time only increases (about n/n-1 times), Basically you can perform tasks. On the other hand, in the case of memory, if the demand is mGB memory, the task itself cannot be executed because the task data cannot fit in the memory unless the mGB memory is allocated. Therefore, if the increase in task execution time can be tolerated, the CPU demand is elastic, and the memory is a non-elastic resource. For elastic resources, even if the total resource demand exceeds the resource capacity, the demand can be accommodated by allocating less resources than the demand, and as a result, the utilization efficiency of other resource types may be improved. There is. However, when there are such multiple resource types, it is difficult to allocate resources satisfying fairness and utilization efficiency in view of the demand type and demand ratio of the resources.

Further, in Non-Patent Document 1, it is assumed that there is no limit on the number of tasks that the user requests to execute, and that there are tasks to be executed as long as resources are allocated. However, in reality, the number of user task requests and the amount of resource requests dynamically change every moment. Non-Patent Document 2 proposes a method of sequentially applying the fair allocation calculated in Non-Patent Document 1 in order to deal with dynamically changing resource requests, but each user makes a resource request only once. Since the model is assumed, there is a problem that it is not applicable when each user generally makes a resource request multiple times in the cloud environment and the resource usage amounts become imbalanced as a result. Also, Non-Patent Document 3 proposes a method for realizing fair and efficient resource distribution by introducing a method called bandwidth reservation when a plurality of users make resource requests successively a plurality of times. However, this document only deals with the case of a single resource, and only assumes the case where the resource is inelastic.

A. Ghodsi, M. Zaharia, B. Hindman, A. ,S. Shenker, and I. Stoica, "Dominant Resource Fairness: Fair Allocation of Multiple Resource Types," Proc. USENIX NSDI'11, Boston, USA, April, 2011. I.Kash, A.D.Procaccia, N.Shah, "No Agent Left Behind: Dynamic Fair Division of Multiple Resources," AAAMAS'2013. S.C.Borst, D.Mitra, "Virtual Partitioning for Robust Resource Sharing: Computational Techniques for Heterogeneous Traffic," IEEE JSAC 1998. http://www.vmware.com/files/jp/pdf/vsp_40_resource_mgmt_ja.pdf vSphere Resource Management Guide, Search April 5, 2017

As described above, when there are multiple resource types, considering resource demand types and demand ratios, it is difficult to allocate resources that satisfy fairness and utilization efficiency in situations where demand is dynamically generated. It was

The present invention has been made in view of the above problems, and in a situation in which a request dynamically occurs in consideration of a demand type and a demand ratio of resources when a plurality of resource types exist, fairness and utilization efficiency It is an object of the present invention to provide a technology that makes it possible to realize resource allocation that satisfies the requirements.

According to the disclosed technology, a resource allocation control system that performs resource allocation control for a user's resource request in an environment in which a plurality of resources are shared by a plurality of users,
Resource request receiving means for receiving a resource request from a user,
A resource request acceptance determining means for determining whether or not to accept the resource request,
The resource request acceptance determination means,
Of the required resources in the resource request, for an elastic resource that is an elastic resource, the sum of the resource request amount of the elastic resource and the total resource usage amount of all users of the elastic resource is the resource capacity. If it does not exceed the threshold given in advance, it is determined that the request regarding the elastic resource can be accepted,
Among the requested resources, for the inelastic resource which is a resource having no elasticity, based on the fair allocation amount, the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource, Is a resource allocation control system that determines that a request for the inelastic resource can be accepted within a range in which the sum of
The resource request acceptance determination means, for the inelastic resource of the request resource,
When the resource usage amount of the inelastic resource of the user does not exceed the fair allocation amount, the sum of the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource is the resource capacity. If it does not exceed the request, it is judged that the request can be accepted, and if the resource usage amount of the user exceeds the fair allocation amount, the value obtained by adding the reserve resource amount to the resource request amount of the inelastic resource and the inelastic resource There is provided a resource allocation control system characterized by determining that a request can be accepted when the sum of the resource usage total of all users does not exceed the resource capacity .

According to the disclosed technology, resource allocation satisfying fairness and utilization efficiency is realized in a situation in which a request dynamically occurs in consideration of resource demand types and demand ratios when there are multiple resource types. It will be possible.

It is a system configuration diagram in an embodiment of the present invention. It is a figure which shows the hardware structural example of the resource allocation control system 100 in embodiment of this invention. It is a figure which shows the specific example of the admission control with respect to an inelastic resource.

Hereinafter, an embodiment of the present invention (this embodiment) will be described with reference to the drawings. The embodiments described below are merely examples, and the embodiments to which the present invention is applied are not limited to the following embodiments.

(System configuration, operation overview)
FIG. 1 shows a system configuration example in this embodiment. As shown in FIG. 1, the system according to this embodiment includes a resource allocation control system 100 and a controlled resource group 200. The resource allocation control system 100 receives a resource request from each user (specifically, the user's terminal) and executes control to allocate each resource in the controlled resource group 200 to each user. That is, the resource allocation control system 100 controls the resource allocation amount with respect to a user's resource request in an environment in which a plurality of users share a plurality of resources.

The control target resource group 200 is, for example, a physical server group that constitutes a virtual machine in the cloud, and is connected to the resource allocation control system 100 via a network. The controlled resource group 200 may be a control device that controls a physical server group that constitutes a virtual machine in the cloud.

The type of each resource in the controlled resource group 200 is not particularly limited, but examples thereof include CPU (number), memory (capacity), and the like. Further, there are two types of resources in the control target resource group 200: a resource having elasticity and a resource having no elasticity.

As shown in FIG. 1, the resource allocation control system 100 includes a resource request reception unit 101, a resource request reception determination unit 102, a resource request prediction unit 103, a fair allocation calculation unit 104, a resource usage amount observation unit 105, and a resource allocation amount calculation unit. With 106. Details of the operation of each functional unit will be described later. The outline of the operation of the resource allocation control system 100 is as follows.

The resource allocation control system 100 calculates the fair allocation amount in advance based on the resource request for each resource for each user. Resource requests are represented, for example, as request vectors. The request vector is described as a demand vector in Non-Patent Document 1 and the like, and represents the resource amount. For example, if a task requires 1 CPU and 4 GB of memory, the request vector of the task can be expressed as (1 CPU, 4 GB). The fair allocation amount can be calculated, for example, based on the method of main resource fairness or resource fairness disclosed in Non-Patent Document 1.

As for the request amount for each resource for each user used for calculating the fair allocation amount, for example, what is dynamically (stochastically) predicted based on the time series model generated from the user's past resource request data is used.

The resource allocation control system 100 determines whether to accept a resource request from a certain user as follows.

For the elastic resource among the requested resources, a request is made unless the sum of the resource request amount of the resource and the total resource usage amount of all users of the resource exceeds a threshold value given in advance based on the resource capacity. Accept.

For resources that do not have elasticity among the requested resources, if the resource usage amount of the user does not exceed the fair allocation amount, the resource request amount of the resource and the resource usage amount of all users of the resource are summed up. If the sum of and does not exceed the resource capacity, the request is accepted, and if the resource usage amount of the user exceeds the fair allocation amount, the spare resource amount calculated separately is added to the resource request amount of the resource. The request is accepted when the sum of the sum of the resource consumption and the total resource usage of all users of the resource does not exceed the resource capacity.

The above-mentioned reserve resource amount can be calculated based on the variation of the request vector (request amount) of each user for the resource. Further, the above-mentioned reserve resource amount may be calculated based on the deviation degree of the actual measurement value from the predicted value dynamically predicted based on the time series model generated from the user's past resource request data described above.

Regarding the user who issued the resource request, the resource request is accepted when it is determined that the request is accepted by all the elastic and inelastic resources.

In response to the resource request received as described above, the resource allocation control system 100 allocates the requested resource to the user as follows.

For elastic resources, if the total number of users requested for the resource exceeds the resource capacity, reallocate the resources so that they become a fair allocation amount, and reallocate resources that do not have elasticity. Is assigned as requested by the user.

<Hardware Configuration of Resource Allocation Control System 100>
The resource allocation control system 100 according to the present embodiment can be realized by causing a computer to execute a program describing the processing content described in the present embodiment. FIG. 2 is a diagram showing a hardware configuration example of the resource allocation control system 100 according to the present embodiment. The resource allocation control system 100 of FIG. 2 includes a drive device 300, an auxiliary storage device 302, a memory device 303, a CPU 304, an interface device 305, a display device 306, an input device 307, etc., which are connected to each other by a bus B. ..

The program that implements the processing in the resource allocation control system 100 is provided by a recording medium 301 such as a CD-ROM or a memory card. When the recording medium 301 storing the program is set in the drive device 300, the program is installed in the auxiliary storage device 302 from the recording medium 301 via the drive device 300. However, it is not always necessary to install the program from the recording medium 301, and the program may be downloaded from another computer via the network. The auxiliary storage device 302 stores the installed program and also stores necessary files and data.

The memory device 303 reads the program from the auxiliary storage device 302 and stores it when an instruction to activate the program is given. The CPU 304 realizes the function related to the resource allocation control system 100 according to the program stored in the memory device 303. The interface device 305 is used as an interface for connecting to a network. The display device 306 displays a GUI (Graphical User Interface) or the like according to a program. The input device 307 includes a keyboard and a mouse, buttons, a touch panel, or the like, and is used to input various operation instructions.

(Operation of Each Functional Unit of Resource Allocation Control System 100)
Hereinafter, the operation of each functional unit in the resource allocation control system 100 shown in FIG. 1 will be described in detail.

<Resource request receiving unit 101>
The resource request reception unit 101 receives a resource request from each user and notifies the resource request reception determination unit 102 of the request. The resource request includes, for example, a request amount for each resource (eg: CPU: 3, memory: 1 GB). The resource request is also notified to the resource request prediction unit 103. The resource request amount of each user changes from moment to moment, and each user issues a resource request every time the resource request is received, and the resource request accepting unit 101 receives the resource request.

<Resource usage monitoring unit 105>
The resource usage amount observation unit 105 observes the usage amount for each resource and each user in the control target resource group 200.

<Resource request acceptance determination unit 102>
The resource request acceptance determination unit 102 determines whether to accept the resource request received from the user. “Accepting a resource request” means allocating the requested resource to the user. However, not all of the requested amount is allocated.

The basic idea of resource request acceptance and resource allocation executed by the resource request acceptance determination unit 102 in the present embodiment is as follows.

The main resource fairness or resource fairness disclosed in Non-Patent Document 1 is set as a desired fair state in a bottleneck state in which the total of resource requests from users exceeds the resource capacity.

In the non-bottleneck state, admission control is performed by using the idea of bandwidth reservation so that the inelastic resource becomes a desirable fair state in the case of a bottleneck state in the future. For elastic resources, resource reallocation is performed so that a desired fair state is achieved after the bottleneck state. Further, even when the elastic resource is in the bottleneck state, it is possible to further accept the resource request by allocating the resource amount allocated per user to the resource amount smaller than the requested resource amount. On the other hand, if the allocated resource amount is too small, the task completion time increases and the occupation time of other resources increases, resulting in an inefficient state. Therefore, with respect to the elastic resource, the admission control is performed so that the allocated resource amount does not become too smaller than the requested resource amount.

At the time of the above-mentioned acceptance determination, the resource request acceptance determination unit 102 uses the reserve resource amount calculated by the resource request prediction unit 103, the resource usage amount observed by the resource usage amount observation unit 105, and the fair allocation calculation unit 104. Based on the calculated fair allocation amount, it is determined whether or not to accept the resource request.

For an elastic resource in the resource request, the resource request reception determination unit 102 adds the resource usage amount of the current user and the resource usage amount of all the users obtained from the resource usage amount observation unit 105. However, it is accepted as long as it does not exceed the predetermined value of the resource capacity. Note that “accept” here means that it is determined that the request for the elastic resource can be accepted. As will be described later, the user's resource request is accepted only when all elastic resource and inelastic resource requests can be accepted.

The above-mentioned multiplication factor is defined by the increase ratio, for the task of the user who makes the resource request, the completion time of which increases compared with the completion time when the requested resource is allocated.

For example, it is assumed that the resource capacity of the relevant resource is 10 and the multiplication factor by which the increase ratio is equal to or less than the regulation is 1.5. Then, assuming that the current usage amount of the resource of all users is 9 and the resource request amount of the relevant user is 3, the sum is 12, which exceeds the resource capacity, but does not exceed 15, so the request is accepted.

For an inelastic resource in the resource request, the resource request acceptance determination unit 102 determines, when the resource usage amount of the user does not exceed the fair allocation amount, the resource request amount of the resource of the user and the resource. If the sum of the total resource usage of all users does not exceed the resource capacity, the request is accepted, and if the resource usage of the user exceeds the fair allocation amount, the resource request amount of the resource is set to the spare resource. If the sum of the sum of the resources and the total resource usage of all users of the resource does not exceed the resource capacity, the request is accepted. It should be noted that “accept” here means that it is determined that the request for the inelastic resource can be accepted.

The resource request acceptance determination unit 102 accepts a user resource request only when all elastic resource and inelastic resource requests included in the request are acceptable. Note that this is an example, and if either the request for the elastic resource or the request for the inelastic resource included in the resource request of the user is acceptable, the request that is permitted is accepted (allocation is performed. ) May be.

A specific example of the admission control for the inelastic resource described above will be described with reference to FIG. FIG. 3A shows the usage amount (A ₁₁ ) of the user 1 for the resource 1 and the usage amount (A ₁₂ ) of the user 2 for the resource 1, and FIG. 3B shows the usage amount of the user 1 for the resource 2 (A 1 ). A ₂₁ ) and the usage amount (A ₂₂ ) of the user 2 are shown. The capacities of resource 1 and resource 2 are C ₁ and C ₂ , respectively.

3A and 3B respectively show a line (indicated by D) in which the resource capacity is the total of the usage amounts of the user 1 and the user 2. Black circles (●) in FIGS. 3A and 3B are fair allocation amounts calculated in advance. The fair share is calculated by the method of fairness of major resources or the method of fairness of assets.

In the state on the line indicated by D in FIGS. 3A and 3B, a resource request (a request in the direction indicated by a black arrow outside the line indicated by D) is received from a user whose resource usage amount is less than the fair allocation amount. I can't. This state is a state that should be avoided.

Therefore, when the resource usage status enters the shaded area indicated by E, the resource usage amount is a fair distribution of the movement in the direction of the white arrow inside the line indicated by D in order to prevent future black arrow events. Admission control is performed to reject the above new resource requests from the user. The resource amount in the shaded area indicated by E corresponds to the above-mentioned reserve resource amount.

<Resource request prediction unit 103>
The resource request prediction unit 103 models the resource request from an appropriate time series model such as an autoregressive model from the past resource request data of each user obtained from the resource request reception unit 101, Is obtained as a demand vector). Further, a value obtained by multiplying an upper P% value of the resource request predicted value by a predetermined ratio N with respect to a predetermined probability P is notified to the resource request acceptance determination unit 102 as a reserve resource amount. For example, when the upper P% value of the resource request predicted value is 10 and N is 1.5, the spare resource amount is 15. This method is also an example of a method of calculating the reserve resource amount based on the variation of the request resource amount of each user (specifically, obtained as a request vector). A method other than the above may be used as a method for calculating the reserve resource amount based on the variation in the request amount of each user for the inelastic resource.

Further, the reserve resource amount may be calculated based on the degree of deviation of the actual measurement value from the resource request predicted value.

<Fair allocation calculation unit 104>
The fair share calculation unit 104 calculates a fair share amount based on the fairness of main resources or a fair share amount based on the asset fairness from the resource request notified from the resource request prediction unit 103, and calculates the calculated fair share amount. The resource request acceptance determination unit 102 is notified. The fair allocation amount for the elastic resource is also notified to the resource allocation amount calculation unit 106.

The fair allocation calculation unit 104 can calculate the fair allocation amount for a resource that does not have elasticity by the method disclosed in Non-Patent Document 1. The fair share allocation for elastic resources is as follows.

The fair allocation calculation unit 104 calculates a fair allocation amount A _ij based on fairness of main resources for elastic resources as a solution to the following optimization problem.

ここでＣ_ｉはリソース種別ｉ（弾力性を持つリソース）の総リソース量である。すなわち、Ｃ_ｉは、弾性リソースに対する要求量のユーザ合計（リソース容量を超えたもの）である。Ａ_ｉｊはリソース種別ｉのユーザｊへの割当リソース量であり、Ｎは総ユーザ数であり、εは予め定められた小数である。ｃ_ｉはリソース種別ｉの単位リソース量であり、ｄ_ｉｊはリソース種別ｉのユーザｊへの割当リソース数である。

Here, C _i is the total resource amount of the resource type i (resource having elasticity). That is, C _i is the total user demand (more than the resource capacity) of elastic resources. A _ij is the amount of resources allocated to the user j of the resource type i, N is the total number of users, and ε is a predetermined decimal number. c _i is the unit resource amount of the resource type i, and d _ij is the number of resources allocated to the user j of the resource type i.

The formula in Subject to is a constraint that it holds for all users j. The first term in the left side of the equation 2 represents the resources for the user j _(A ij / _{C i)} is the maximum value of _(A ij / _{C i),} the second term, (A for each user _ij / _{C i)} it has a value that the sum of the total users divided by N values _(a ij / _{C i)} in the resource becomes maximum.

The above calculation is a calculation to obtain A _ij that maximizes the sum of the total users and total resources of A _ij under the condition of Expression 2.

Further, the fair share calculation unit 104 calculates a fair share amount A _ij based on the fairness of assets for elastic resources as a solution to the following optimization problem.

ここでＣ_ｉはリソース種別ｉ（弾力性を持つリソース）の総リソース量であり、Ａ_ｉｊはリソース種別ｉのユーザｊへの割当リソース量である。Ｎは総ユーザ数であり、εは予め定められた小数である。

Here, C _i is the total resource amount of the resource type i (resource having elasticity), and A _ij is the resource amount allocated to the user j of the resource type i. N is the total number of users, and ε is a predetermined decimal number.

The formula in Subject to is a constraint that it holds for all users j. The first term on the left-hand side of Equation 5 is the sum of (A _ij /C _i ) total resources, and the second term is the sum of (A _ij /C _i ) total users and total resources divided by N. Indicates a value.

The above calculation is a calculation to obtain A _ij that maximizes the sum of the total users and total resources of A _ij under the condition of Expression 5.

<Resource allocation amount calculation unit 106>
When the usage amount of the resource reaches the resource capacity, the resource allocation amount calculation unit 106 performs resource allocation (re-allocation) based on the fair allocation amount notified from the fair allocation calculation unit 104. )I do. For resources that do not have elasticity, resource allocation is performed as requested by the user. “Perform resource allocation” means that, for example, the resource allocation amount calculation unit 106 notifies the control target resource group 200 of the allocation amount of each allocation resource to the user, and the control target resource group 200 is allocated with the resource of the allocation amount. Is to be secured by the user.

According to the technology according to the present embodiment described above, fairness and utilization efficiency are improved in a situation where a request dynamically occurs in consideration of a demand type and a demand ratio of resources when there are multiple resource types. Expected to meet the resource allocation.

(Summary of Embodiments)
As described above, according to the present embodiment, a resource allocation control system that performs resource allocation control for a user's resource request in an environment in which multiple resources are shared by multiple users, and receives a resource request from a user. Resource request acceptance determining means and resource request acceptance determining means for determining whether or not to accept the resource request, wherein the resource request acceptance determining means is an elastic resource among the request resources in the resource request. If the sum of the resource demand of the elastic resource and the total resource usage of all users of the elastic resource does not exceed the threshold given in advance based on the resource capacity, A request regarding a resource is determined to be acceptable, and for the inelastic resource that is a resource having no elasticity among the requested resources, the resource request amount of the inelastic resource and the inelastic resource are based on the fair allocation amount. There is provided a resource allocation control system characterized by determining that a request relating to the inelastic resource can be accepted within a range in which the sum of the total resource usage of all users does not exceed the resource capacity.

The resource request acceptance unit 101 is an example of a resource request acceptance unit, and the resource request acceptance determination unit 102 is an example of a resource request acceptance determination unit.

The resource request acceptance determination means, for example, when the resource usage amount of the inelastic resource of the user does not exceed a fair allocation amount with respect to the inelastic resource among the requested resources, If the sum of the resource request amount and the total resource usage amount of all users of the inelastic resource does not exceed the resource capacity, it is determined that the request can be accepted, and the resource usage amount of the user exceeds the fair allocation amount. In this case, it is determined that the request can be accepted when the sum of the resource request amount of the inelastic resource plus the reserve resource amount and the total resource usage amount of all users of the inelastic resource does not exceed the resource capacity. ..

The resource allocation control system may include a calculation unit that calculates the reserve resource amount based on a variation in a request amount of each user for the inelastic resource. In addition, the resource allocation control system obtains a predicted value that is a stochastic prediction of a request amount for each resource for each user based on a time series model generated from the user's past resource request data, and calculates the spare resource amount. A calculation means for calculating based on the degree of deviation of the actually measured value from the predicted value may be provided. The resource request prediction unit 103 is an example of calculation means.

Further, the resource allocation control system performs re-allocation calculation for the elastic resource so as to be a fair allocation amount when the user total of the resource request amount of the elastic resource exceeds the resource capacity, and by the calculation. It may further include an allocation unit that allocates the obtained amount of resources to the user and allocates resources to the inelastic resource as requested by the user. The fair allocation calculation unit 104 and the resource allocation amount calculation unit 106 are examples of allocation means.

Although the present embodiment has been described above, the present invention is not limited to this particular embodiment, and various modifications and changes can be made within the scope of the gist of the present invention described in the claims. It is possible.

100 Resource Allocation Control System 200 Controlled Resource Group 101 Resource Request Accepting Unit 102 Resource Request Acceptance Determining Unit 103 Resource Request Predicting Unit 104 Fair Allocation Calculating Unit 105 Resource Utilization Observing Unit 106 Resource Allocation Calculating Unit 300 Drive Device 301 Recording Medium 302 Auxiliary storage device 303 Memory device 304 CPU
305 Interface device 306 Display device 307 Output device

Claims (6)

In an environment in which multiple resources are shared by multiple users, a resource allocation control system that performs resource allocation control for user resource requests,
Resource request receiving means for receiving a resource request from a user,
A resource request acceptance determining means for determining whether or not to accept the resource request,
The resource request acceptance determination means,
Of the required resources in the resource request, for an elastic resource that is an elastic resource, the sum of the resource request amount of the elastic resource and the total resource usage amount of all users of the elastic resource is the resource capacity. If it does not exceed the threshold given in advance, it is determined that the request regarding the elastic resource can be accepted,
Among the requested resources, for the inelastic resource which is a resource having no elasticity, based on the fair allocation amount, the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource, Is a resource allocation control system that determines that a request for the inelastic resource can be accepted within a range in which the sum of
The resource request acceptance determination means, for the inelastic resource of the request resource,
When the resource usage amount of the inelastic resource of the user does not exceed the fair allocation amount, the sum of the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource is the resource capacity. If it does not exceed the request, it is judged that the request can be accepted, and if the resource usage amount of the user exceeds the fair allocation amount, the value obtained by adding the reserve resource amount to the resource request amount of the inelastic resource and the inelastic resource A resource allocation control system characterized by determining that a request can be accepted when the sum of the resource usage total of all users does not exceed the resource capacity . The resource allocation control system according to claim 1 , further comprising: a calculation unit that calculates the reserve resource amount based on a variation in a request amount of each user for the inelastic resource. The requested amount for each resource for each user is calculated stochastically based on the time series model generated from the user's past resource request data, and the predicted value is calculated stochastically, and the spare resource amount is used as the deviation degree of the actual measured value from the predicted value. The resource allocation control system according to claim 1 , further comprising: a calculating unit that calculates based on the above. For the elastic resource, when the user total of the resource request amount of the elastic resource exceeds the resource capacity, the reallocation calculation is performed so as to become the fair allocation amount, and the resource of the amount obtained by the calculation is calculated by the user. The resource allocation control system according to any one of claims 1 to 3 , further comprising: an allocation unit that allocates the resource to the inelastic resource as requested by the user. In an environment in which a plurality of resources are shared by a plurality of users, a resource allocation control method executed by a resource allocation control system that performs resource allocation control for user resource requests,
A resource request receiving step of receiving a resource request from a user,
A resource request acceptance determining step of determining whether to accept the resource request,
In the resource request acceptance determination step, the resource allocation control system,
Of the required resources in the resource request, for an elastic resource that is an elastic resource, the sum of the resource request amount of the elastic resource and the total resource usage amount of all users of the elastic resource is the resource capacity. If it does not exceed the threshold given in advance, it is determined that the request regarding the elastic resource can be accepted,
Among the requested resources, for the inelastic resource which is a resource having no elasticity, based on the fair allocation amount, the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource, Is a resource allocation control method that determines that a request for the inelastic resource can be accepted within a range in which the sum of
In the resource request acceptance determination step, the resource allocation control system, for the inelastic resource of the request resource,
When the resource usage amount of the inelastic resource of the user does not exceed the fair allocation amount, the sum of the resource request amount of the inelastic resource and the total resource usage amount of all users of the inelastic resource is the resource capacity. If it does not exceed the request, it is judged that the request can be accepted, and if the resource usage amount of the user exceeds the fair allocation amount, the value obtained by adding the reserve resource amount to the resource request amount of the inelastic resource and the inelastic resource If the sum of the total resource usage of all users of the resource does not exceed the resource capacity, it is determined that the request can be accepted.
A resource allocation control method characterized by the above . A program for causing a computer to function as each unit in the resource allocation control system according to any one of claims 1 to 4 .

Images