KR20180029635A - Apparatus for predicting resource and cloud server - Google Patents

Abstract

The present invention provides a resource prediction device and a cloud server, which predict a demand amount of a user from a cloud-based database and determine an amount of a resource to be next provided by using a predicted result value. The device for predicting a resource demand amount with respect to a virtual resource comprises: a penalty cost deriving part deriving a first server penalty cost based on a response time in which requests from a plurality of user terminals are processed by at least one virtual machine pre-created in the cloud server, and deriving a second service penalty cost based on a resource sharing amount including a CPU sharing amount provided through the virtual resource of the cloud server and a memory sharing amount; a resource predicting part comparing the first service penalty cost and the second service penalty cost, and deriving a required resource prediction value; and an allocation requesting part requesting a dynamic allocation of the virtual resource to the cloud server based on the resource prediction value.

Description

[0001] APPARATUS FOR PREDICTING RESOURCE AND CLOUD SERVER [0002]

The present invention relates to a resource predicting apparatus and a cloud server.

Cloud Computing refers to the environment in which users can use the Internet at any time without having to install the necessary software on their computers and easily share the data with various information and communication devices at the same time. Cloud computing can be used to process big data, and distributed processing through multiple servers is essential for processing big data.

With respect to the technology for distributing data in such a cloud computing environment, Korean Patent Registration No. 10-1195123 discloses a cloud computing network system and a method for distributing the file.

Service providers can establish service level agreements (SLAs) with their customers to provide a cloud computing environment and provide customers with predefined levels of service based on service level agreements.

Conventionally, according to the static service level agreement, the same service was provided regardless of the circumstances of the network, and thus, the service level penalty cost was increased.

A resource prediction apparatus and a cloud server that predict a user's demand amount from a cloud-based database and determine the amount of resources to be provided next using predicted result values. The present invention provides a resource forecasting device and a cloud server that can reduce the cost of services by controlling the resource sharing amount of the cloud-based database through accurate resource prediction and using the necessary resources. A resource prediction device and a cloud server capable of satisfying a user's QoS by dynamically adjusting a resource sharing amount. It is to be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

According to an aspect of the present invention, there is provided a method for managing a service of a first service based on a response time in which a request from a plurality of user terminals is processed by at least one virtual machine previously created in a cloud server, A penalty cost deriving unit for deriving a penalty cost and deriving a second service penalty cost based on a resource sharing amount including a CPU share amount and a memory share amount provided through a virtual resource of the cloud server; A resource predicting unit for comparing a first service penalty cost and a second service penalty cost to obtain a required resource predicted value and an allocation request unit for requesting dynamic allocation of the virtual resource to the cloud server based on the resource predicted value A resource prediction device.

According to an example, the resource prediction value may include a first direction value corresponding to the CPU sharing amount and a second direction value corresponding to the memory sharing amount. The resource sharing amount may further include a number of the virtual machines, and the resource estimation value may further include a third direction value corresponding to the number of the virtual machines.

According to an example, the penalty cost derivation unit may derive the first service penalty cost based on whether a response time to a request processed by a virtual machine corresponding to each of the plurality of user terminals exceeds a reference value. The penalty cost derivation unit may derive the first service penalty cost by applying different weights according to the virtual machine processing the request.

According to an example, the resource predicting unit may apply the CPU sharing amount, the memory sharing amount, and the number of virtual machines as a component of a three-dimensional array, and calculate the first service penalty cost and the The second direction value, and the third direction value from the 2-service penalty cost.

According to an example, the resource predicting unit may adjust the second direction value and adjust the first direction value when the memory sharing amount reaches a limit. The resource predicting unit adjusts the first direction value to adjust the third direction value when the CPU sharing amount reaches the limit, and the allocation request unit adds the virtual machine to the virtual machine according to the adjusted third direction value. To the cloud server.

According to one example, the penalty cost derivation unit may derive an added infrastructure cost based on at least one of the number and performance of the virtual machines, and add the infrastructure cost to the second service penalty cost. And the infrastructure cost may be changed according to the third direction value.

According to an example, the penalty cost derivation unit may derive a delay cost based on a delay time generated according to the additional generation of the virtual machine when a virtual machine is additionally generated according to the third direction value, 2 service penalties.

According to another embodiment of the present invention, there is provided a virtual machine system including a virtual machine generation unit for generating at least one virtual machine from a virtual resource based on a predetermined service level agreement, a processing unit for processing a request from a plurality of user terminals via the generated virtual machine And an allocation unit for dynamically allocating the virtual resource based on a resource prediction value received from the resource prediction apparatus, wherein the resource prediction value is calculated based on a response time in which a request from the plurality of user terminals is processed by the virtual machine Based on a derived second service penalty cost derived based on a derived resource sharing amount including a first service penalty cost and a CPU sharing amount and a memory sharing amount provided through virtual resources of the cloud server, A cloud server that is derived by the prediction device can be provided.

The above-described task solution is merely exemplary and should not be construed as limiting the present invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and the detailed description of the invention.

According to any one of the above-described tasks, a resource prediction device and a cloud server for predicting a demand amount of a user from a cloud-based database and determining the amount of a resource to be provided next using the predicted result value, . It is possible to provide a resource predicting device and a cloud server that can reduce the cost required for a service by controlling the amount of resource sharing of the cloud-based database through accurate resource prediction and using as much resources as necessary. By dynamically adjusting the resource sharing amount, it is possible to provide a resource prediction device and a cloud server that can satisfy a user's QoS.

1 is a configuration diagram of a resource prediction system according to an embodiment of the present invention.
2 is a configuration diagram of a resource predicting apparatus according to an embodiment of the present invention.
3A and 3B are exemplary diagrams illustrating a process of deriving a first service penalty cost according to an embodiment of the present invention.
4A and 4B are exemplary diagrams illustrating a second service penalty cost that reflects a delay cost according to an embodiment of the present invention.
FIG. 5 is an exemplary diagram illustrating a process of deriving a resource prediction value according to an embodiment of the present invention. Referring to FIG.
6 is an exemplary diagram illustrating a cost definition table in accordance with an embodiment of the present invention.
7 is an exemplary diagram comparing an existing static service level agreement technique and a dynamic service level agreement technique in accordance with an embodiment of the present invention.
8 is a flowchart of a method for predicting a resource requirement for a virtual resource in a resource prediction apparatus according to an embodiment of the present invention.
9 is a configuration diagram of a cloud server according to an embodiment of the present invention.
10 is a flowchart of a method of sharing virtual resources in cooperation with a resource predicting device in a cloud server according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "including" an element, it is to be understood that the element may include other elements as well as other elements, And does not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

In this specification, the term " part " includes a unit realized by hardware, a unit realized by software, and a unit realized by using both. Further, one unit may be implemented using two or more hardware, or two or more units may be implemented by one hardware.

In this specification, some of the operations or functions described as being performed by the terminal or the device may be performed in the server connected to the terminal or the device instead. Similarly, some of the operations or functions described as being performed by the server may also be performed on a terminal or device connected to the server.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a resource prediction system according to an embodiment of the present invention. Referring to FIG. 1, a plurality of user terminals 110, a resource prediction device 120, and a cloud server 130 may be included. The plurality of user terminals 110, the resource prediction device 120 and the cloud server 130 shown in FIG. 1 illustratively show the components that can be controlled by the resource prediction system 1.

Each component of the resource prediction system 1 of Fig. 1 is generally connected via a network. For example, as shown in FIG. 1, the resource prediction apparatus 120 may be connected to the main server 130 at the same time or at a time interval via a network.

The network refers to a connection structure in which information can be exchanged between each node such as terminals and servers. An example of such a network is Wi-Fi, Bluetooth, Internet, LAN (Local Area Network ), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), 3G, 4G, 5G, LTE and the like.

The plurality of user terminals 110 may request the cloud server 130 to process a query generated in the terminal. The plurality of user terminals 110 may receive a response to the processed request through the virtual machine created by the cloud server 130.

The resource predicting device 120 calculates a first service penalty cost based on a response time in which a request from a plurality of user terminals 110 is processed by at least one virtual machine created in the cloud server 130, Can be derived. For example, the resource predicting device 120 may derive a first service penalty cost based on whether the response time for a request processed by the virtual machine corresponding to each of the plurality of user terminals 110 exceeds a reference value . The resource predicting device 120 may derive the first service penalty cost by applying different weights according to the virtual machine processing the request.

The resource predicting device 120 can derive the second service penalty cost based on the resource sharing amount including the CPU share amount, the memory share amount, and the number of virtual machines provided through the virtual resources of the cloud server 130 . The resource prediction device 120 may derive the added infrastructure cost based on at least one of the number and performance of the virtual machines and add the infrastructure cost to the second service penalty cost. When the virtual machine is additionally generated according to the predicted direction value, the resource predicting device 120 derives the delay cost based on the delay time generated according to the additional generation of the virtual machine, . ≪ / RTI >

The resource predicting device 120 can compare the first service penalty cost and the second service penalty cost to derive a required resource predicted value. The resource prediction unit 120 may include a first direction value corresponding to the amount of CPU sharing, a second direction value corresponding to the amount of memory sharing, and a third direction value corresponding to the number of virtual machines, The amount of sharing, the amount of memory sharing, and the number of virtual machines can be applied as components of a three-dimensional array. For example, the resource prediction apparatus 120 may derive a first direction value, a second direction value, and a third direction value from the first service penalty cost and the second service penalty cost using the three-dimensional array. At this time, the resource predicting device 120 adjusts the first direction value when the memory sharing amount reaches the limit by adjusting the second direction value, and when the CPU sharing amount reaches the limit by adjusting the first direction value, The third direction value can be adjusted. In addition, the resource predicting device 120 may change the infrastructure cost according to the third direction value.

The resource predicting device 120 may request dynamic allocation of virtual resources to the cloud server 130 based on the resource predicted value. For example, the resource prediction device 120 may request the cloud server 130 to generate additional virtual machines by adjusting the third direction value when the CPU share amount reaches the limit by adjusting the first direction value .

The cloud server 130 may create at least one or more virtual machines from virtual resources based on predetermined service level agreements. The cloud server 130 may process the requests from the plurality of user terminals 110 through the generated virtual machine.

The cloud server 130 can dynamically allocate a virtual resource based on the resource prediction value received from the resource prediction device 120. [ For example, the resource predicting device 120 adjusts the second direction value to adjust the second direction value to the third direction value based on at least one of the cases where the memory sharing amount reaches the limit and the CPU share amount reaches the limit by adjusting the first direction value. When adjusting the direction value, the cloud server 130 may additionally generate a virtual machine according to the adjusted third direction value. At this time, the cloud server 130 can dynamically allocate the virtual resources considering the added infrastructure cost based on at least one of the number and performance of the virtual machines. In addition, when the number of virtual machines is additionally generated according to the third direction value, the cloud server 130 dynamically allocates the virtual resources considering the delay cost added based on the delay time generated according to the additional generation of the virtual machine Can be assigned.

2 is a configuration diagram of a resource predicting apparatus according to an embodiment of the present invention. Referring to FIG. 2, the resource prediction apparatus 120 may include a penalty cost derivation unit 210, a resource prediction unit 220, and an allocation request unit 230.

The penalty cost deriving unit 210 may calculate the first service penalty cost and the second service penalty cost necessary for calculating the average SLA cost.

For example, the penalty cost deriving unit 210 may determine whether or not the first service disadvantage (e.g., the first service disadvantage) based on the response time at which the request from the plurality of user equipments 110 is processed by the at least one virtual machine created in the cloud server 130 Cost can be deduced. The process of deriving the first service penalty cost will be described in detail with reference to FIGS. 3A and 3B.

3A and 3B are exemplary diagrams illustrating a process of deriving a first service penalty cost according to an embodiment of the present invention.

3A is an exemplary diagram illustrating gain cost and penalty costs according to response time in accordance with an embodiment of the present invention. 3A, the penalty cost derivation unit 210 may calculate a penalty cost cost 210 based on whether a response time to a request processed by a virtual machine corresponding to each of the plurality of user terminals 110 exceeds a reference value, Can be derived. For this, the penalty cost deriving unit 210 may calculate the gain cost 300 and the penalty cost 310 based on the response time X _r .

[Equation 1]

Equation (1) may be a formula for deriving the first service penalty cost value. Referring to Equation (1), the penalty cost derivation unit 210 grants the gain cost 300 if the measurement value of the response time is smaller than the reference value X _r specified in the service level agreement (SLA, Service Level Agreement) If the measure of time is greater than the reference value X _r specified in the Service Level Agreement (SLA), penalty cost 310 may be awarded. The penalty cost deriving unit 210 can be defined as Equation 1 by expressing the sum of the gain cost 300 and the penalty cost 310 as 1.

&Quot; (2) "

Equation (2) may be a formula for deriving the average penalty cost value. Referring to Equation (2), the penalty cost derivation unit 210 may define the average service penalty cost value by substituting the first service penalty cost value derived from Equation (1). The penalty policy deriving unit 210 calculates the penalty cost value of the first service derived from Equation 1 every time a request is made from a plurality of user terminals and divides the calculated result value by the total number of requests, The penalty cost value can be derived. For example, if ten requests originated from the user terminal 110, but one of them could not be processed, an average service penalty cost value of 1/10 = 0.1 could be derived. The penalty policy deriving unit 210 may compare the average service penalty cost value derived through Equation (2) with the average service penalty cost value calculated as the resource sharing amount of the database.

FIG. 3B is an exemplary diagram illustrating a process of applying weights differently according to a virtual machine according to an exemplary embodiment of the present invention. Referring to FIG. Referring to FIG. 3B, the penalty cost deriving unit 210 may derive the first service penalty cost by applying different weights according to the performance of the virtual machine processing the request or the response time in which the request is processed. For example, the better the performance of the virtual machine, the faster the request can be processed. Therefore, the weights can be differently applied to each request by dividing into Slow (320), Normal (321), and Fast (322) according to the response time in which the request is processed.

&Quot; (3) "

Equation (3) may be a formula defining a first service penalty cost function to which weights are applied. Referring to Equation (3), the penalty cost derivation unit 210 may calculate the i-th average service penalty cost at intervals of time t using the first service penalty cost function.

&Quot; (4) "

Equation (4) may be a formula of the first service penalty cost to which the entire weight is applied. Referring to Equation (4), if the number of N user terminals 110 is given, the penalty cost derivation unit 210 may calculate the first service penalty cost by applying the entire weight.

Returning to FIG. 2, the penalty-cost derivation unit 210 calculates a penalty cost based on a resource sharing amount including a CPU sharing amount, a memory sharing amount, and a number of virtual machines provided through the virtual resources of the cloud server 130, Penalty costs can be derived.

&Quot; (5) "

Equation (5) may be a formula for deriving the second service penalty cost. Referring to Equation (5), the penalty cost derivation unit 210 obtains the second service penalty cost using the resource sharing amount including the CPU sharing amount, the memory sharing amount, the copy number of the virtual machine, Can be calculated. According to the calculation result using Equation (4), when the CPU share amount, the memory sharing amount, and the number of replicated virtual machines increase, the average query cost increases and the request input rate and system load increase, The cost increases. For example, according to Equation (4), in the case of 1% CPU sharing amount, the cost of about 0.5% is reduced, and when the number of copies of the virtual machine is increased by one, the cost of 4.3% can be reduced.

The penalty cost deriving unit 210 may derive the added infrastructure cost based on at least one of the number and performance of the virtual machines and add the infrastructure cost to the second service penalty cost. The performance and number of virtual machines affects infrastructure costs, and the addition and deletion of virtual machines affect service level agreements (SLAs).

&Quot; (6) "

Equation (6) may be a formula defining the infrastructure cost reflecting the number of virtual machines (VM (t)).

&Quot; (7) "

Equation (7) may be a formula defining the infrastructure cost that reflects the change. Referring to FIG. 7, when the provisioning is performed in the next virtual machine, the infrastructure cost may be changed, and the penalty cost derivation unit 210 may reflect the previous infrastructure cost and the addition or deletion of the virtual machine to the infrastructure cost can do. For example, D ₃ is 1 if a virtual machine is added, D ₃ is 0 if the virtual machine is maintained, and D ₃ is -1 if the virtual machine is deleted.

If the virtual machine is additionally generated according to the third direction value corresponding to the number of virtual machines derived by the resource predicting unit 220, the penalty cost deriving unit 210 calculates the penalty cost of the virtual machine based on the delay time And the delay cost can be added to the second service penalty cost. For example, when a virtual machine is additionally created, the additional virtual machine is not directly associated with the existing virtual machine, and a delay time from the assignment of the IP address to the construction of the database is generated. Accordingly, if a data processing request is received from a database of an existing virtual machine during a delay time, a processing failure occurs, and a delay cost is incurred. The delay cost will be described in detail with reference to FIGS. 4A and 4B.

4A and 4B are exemplary diagrams illustrating a second service penalty cost that reflects a delay cost according to an embodiment of the present invention.

4A is an exemplary diagram illustrating a service penalty cost according to time of a virtual machine according to an embodiment of the present invention. Referring to FIG. 4A, the average service penalty cost 410 according to the time interval 400 is calculated by using the time interval 400 as the x-axis and the average service penalty cost 410 as the y-axis, The degree of processing can be indicated.

For example, the database of the virtual machine can see that the service penalty cost 410 increases from time interval 400 '80'. At this time, the increased cost of service penalty may be caused by reaching the limit of data processing (420) by not processing the requested data in the virtual machine's database.

The database of the virtual machine can know that the service penalty cost is lowered while the virtual machine is linked with the created virtual machine at time interval 400 '95'.

&Quot; (8) "

Equation (8) may be a formula for deriving a delay time in the additional creation of a virtual machine. Referring to Equation (8), the penalty cost derivation unit 210 can derive a delay time according to the additional creation of a virtual machine (when D ₃ = 1) in order to prevent an increase in the service penalty cost.

4B is an exemplary diagram illustrating a distribution of differences in service penalty costs according to a delay time of a virtual machine according to an exemplary embodiment of the present invention. Referring to FIG. 4B, the penalty cost derivation unit 210 calculates a penalty cost cost for the second service penalty cost 440 with respect to the time 435 during which the delay time occurs during the total time (T, 430) 450) can be added.

Returning to FIG. 2, the resource predicting unit 220 may compare the first service penalty cost and the second service penalty cost to derive a required resource predicted value. The resource prediction value may include a first direction value corresponding to a CPU share amount, a second direction value corresponding to a memory sharing amount, and a third direction value corresponding to the number of virtual machines. A process of deriving a direction value using a three-dimensional array will be described in detail with reference to FIG.

FIG. 5 is an exemplary diagram illustrating a process of deriving a resource prediction value according to an embodiment of the present invention. Referring to FIG. Referring to FIG. 5, the resource predicting unit 220 applies a CPU sharing amount, a memory sharing amount, and a number of virtual machines as constituent elements of a three-dimensional array, and calculates a first service penalty cost and a second service penalty cost using a three- The first direction value 500, the second direction value 510, and the third direction value 520 can be derived from the service penalty cost. In this case, the resource predicting unit 220 may adjust the first direction value 500 when the memory sharing amount reaches the limit by adjusting the second direction value 510, and may adjust the first direction value 500 So that the third direction value 520 can be adjusted when the CPU share amount reaches the limit.

&Quot; (9) "

Equation (9) may be an average penalty cost formula based on the amount of database resource sharing of the virtual machine. Referring to Equation (9), the resource predicting unit 220 can assign the derived result value to AC (i, t) in Equation (9) when the first service penalty cost is derived according to the response time in Equation have. VM (t-1) in Equation (9) is the number of virtual machines, and R (i, t) may be a representation of the requested throughput of the virtual machine.

Resource prediction unit 220 includes a CPU share amounts, memory sharing amount, sure of a resource sharing amount containing the number of the virtual machine, the first direction value by using the resource-sharing capacity (500, d _i1), the second direction value (510, d _i2 ) can be derived.

At this time, the resource predicting unit 220 can satisfy the average service penalty cost derived from Equation (3) by adjusting the first direction value 500, d _i1 , and the second direction value 510, d _i2 . In addition, the resource predicting unit 220 can preferentially adjust the second direction values 510 and d _i2 , which are parameter change values that can increase or decrease the size of the database shared memory of the virtual machine, If the average service penalty cost of Equation (3) can not be satisfied, the cloud server 130 may request the allocation request unit 230 based on the adjusted first direction value and the second direction value by adjusting the first direction value, Quot; virtual resource "

&Quot; (10) "

Equation (10) may be an average penalty cost formula based on the database resource sharing amount of the virtual machine as an expanded mathematical expression with respect to the number of virtual machines from Equation (9). Referring to Equation (10), the resource predicting unit 220 may apply a CPU sharing amount, a memory sharing amount, and a number of virtual machines as a component of a three-dimensional array.

Since the resource predicting unit 220 knows the resource sharing amount including the CPU sharing amount, the memory sharing amount, and the number of virtual machines, the resource predicting unit 220 calculates the first direction value d _i1 , the second direction value d _i2 ) and the third direction value ( _di3 ). In this case, the resource predicting unit 220 may calculate the average service penalty cost (d i) derived from Equation (3) by adjusting the first direction value d _i1 , the second direction value d _i2 and the third direction value d _i3 , Can be satisfied. In addition, the resource predicting unit 220 can preferentially adjust the second direction value d _i2 , which is a parameter change value that can increase or decrease the size of the database shared memory of the virtual machine, When the amount reaches the limit, if the average service penalty cost of Equation (3) is not satisfied, the first direction value can be adjusted. In addition, the resource predicting unit 220 may adjust the third direction value when the CPU share amount reaches the limit by adjusting the first direction value. In this case, the allocation request unit 23 may request the cloud server 130 to generate additional virtual machines according to the adjusted third direction value.

&Quot; (11) "

Equation (11) may be used to change the service level agreement cost using the average service penalty cost. Referring to Equation (11), by assigning the average service penalty cost to Equation (11), the service level agreement cost value is changed. At this time, the response time of the database constructed on the virtual machine changes according to the performance of the virtual machine, and the weight w (i) of the virtual machine changes depending on the response time. Weights are given to virtual machines with fast response times according to the response time of the cloud-based database, while those with slow response times are given low weights, and the higher the weight, the greater the contribution to resource sharing .

&Quot; (12) "

Equation (12) may be a formula for deriving the next direction value. Referring to Equation (12), when the average service level cost value is derived through Equation (11), the resource predicting unit 220 can derive the next direction value using the dynamically changed average service level agreement cost value.

For example, if the CPU sharing amount is 20%, the memory sharing amount is 1024MB, and the three-dimensional array value for predicting the next resource is a first direction value ( _di1 ) = 10 and a second direction value ( _di2 ) = -100 , The resource predicting unit 220 can predict the next resource predicted value as the CPU sharing amount 30% and the memory sharing amount 924 MB. The resource predicting unit 220 can derive a next direction value based on the predicted resource sharing amount.

The resource predicting unit 220 may change the infrastructure cost according to the third direction value.

&Quot; (13) "

Equation (13) may be used to derive the next direction value using the infrastructure cost. When requesting additional creation of a virtual machine by resource prediction, the infrastructure can provide various performance virtual machines. Since the virtual machines that can be created have different costs depending on their performance, parameters are needed to reflect them. Hence, Equation (13) defines the parameter as an infrastructure cost and reflects it in the resource prediction.

Also, the resource predicting unit 220 can derive the next resource prediction direction according to the delay cost according to the delay time.

&Quot; (14) "

Equation (14) may be a formula for deriving the next direction value using the delay cost. The delay cost can define the time from creation request to virtual machine to database connection. In other words, the delay time can be prevented from being delayed according to the delay time of the database-based database expansion by reflecting the delay cost into the resource prediction.

Referring back to FIG. 2, the allocation request unit 230 may request the dynamic allocation of virtual resources to the cloud server 130 based on the resource prediction value. For example, if the third direction value is adjusted according to the resource prediction value, the allocation request unit 230 may request the cloud server 130 to additionally generate a virtual machine.

6 is an exemplary diagram illustrating a cost definition table in accordance with an embodiment of the present invention. Referring to FIG. 6, the resource prediction apparatus 120 may charge a cost differently depending on the performance of the virtual machine.

The resource predicting apparatus 120 may calculate a service penalty cost 620 including a number of virtual machines 610, a first service penalty cost and a second service penalty cost, an infrastructure cost 630, an average number of virtual machines 640, The total service level agreement cost 650 can be derived.

At this time, the lower the total service level agreement cost, the lower the cost, and the resource predicting device 120 can dynamically change the number of virtual machines in consideration of the total service level agreement cost. For example, if the cost c is 'c = 1' (600), it represents a relatively low price and low performance virtual machine. In case of 'c = 3' (601) Machine. At this time, the resource predicting device 120 can lower the total service level agreement cost by using the dynamic service level agreement method, thereby dynamically changing the resource sharing amount of the virtual machine database and the number of virtual machines So that the performance can be improved at a low cost.

7 is an exemplary diagram comparing an existing static service level agreement technique and a dynamic service level agreement technique in accordance with an embodiment of the present invention. Referring to FIG. 7, when the number of virtual machines increases, the service level agreement penalty cost whose resource sharing amount is adjusted by the dynamic service level agreement 720 proposed in the present invention exceeds the cost of the existing static service level agreement 710 It can be seen that it is less.

8 is a flowchart of a method for predicting a resource requirement for a virtual resource in a resource prediction apparatus according to an embodiment of the present invention. A method for predicting a resource requirement for a virtual resource performed in the resource predicting apparatus 120 according to the embodiment shown in FIG. 8 is performed in a time-series manner in the resource predicting system 1 according to the embodiment shown in FIG. 1 / RTI > Accordingly, the present invention is also applied to a method of predicting a resource requirement for a virtual resource performed by the resource predicting apparatus 120 according to the embodiment shown in FIGS. 1 to 7 even if omitted from the following description.

In step S810, the resource predicting device 120 calculates a first service penalty cost (i.e., a first service penalty cost) based on the response time at which the request from the plurality of user terminals 110 is processed by the at least one virtual machine created in the cloud server 130 penalty cost can be derived.

In step S820, the resource prediction device 120 may derive a second service penalty cost based on a resource sharing amount including a CPU share amount and a memory share amount provided through the virtual resources of the cloud server 130. [

In step S830, the resource prediction apparatus 120 may compare the first service penalty cost and the second service penalty cost to derive a required resource prediction value.

In step S840, the resource prediction apparatus 120 may request the dynamic allocation of the virtual resource to the cloud server 130 based on the resource prediction value.

In the above description, steps S810 to S840 may be further divided into further steps or combined into fewer steps, according to an embodiment of the present invention. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

9 is a configuration diagram of a cloud server according to an embodiment of the present invention. 9, the cloud server 130 may include a virtual machine generation unit 910, a processing unit 920, and an allocation unit 930.

The virtual machine creation unit 910 may create at least one virtual machine from virtual resources based on a predetermined service level agreement. The predefined service level agreements may specify the requirements of the virtual resources. The virtual machine generation unit 910 may analyze at least one service level agreement to generate at least one or more virtual machines.

The processing unit 920 can process requests from the plurality of user terminals 110 through the generated virtual machine.

The allocation unit 930 can dynamically allocate a virtual resource based on the resource prediction value received from the resource prediction apparatus 120. [ The resource forecast value is calculated based on the first service penalty cost derived based on the response time when the request from the plurality of user terminals 110 is processed by the virtual machine and the CPU share amount provided through the virtual resource of the cloud server 130, The resource predicted value is derived by the resource predicting device 120 based on the second service penalty cost derived based on the resource sharing amount including the share amount. The resource predicted value is a first direction value corresponding to the CPU share amount, And a third direction value corresponding to the number of virtual machines.

For example, if the resource predicting device 120 adjusts the second direction value to reach the limit for the amount of memory sharing and adjusts the first direction value to reach the limit for the CPU share amount, If the third direction value is adjusted based on the third direction value, the assigning unit 930 can additionally generate a virtual machine according to the adjusted third direction value. At this time, the allocating unit 930 can dynamically allocate virtual resources considering the added infrastructure cost based on at least one of the number of virtual machines and the performance. In addition, when the number of virtual machines is additionally generated according to the third direction value, the allocating unit 930 dynamically allocates the virtual resources considering the delay cost added based on the delay time generated according to the additional generation of the virtual machine Can be assigned.

10 is a flowchart of a method of sharing virtual resources in cooperation with a resource predicting device in a cloud server according to an embodiment of the present invention. A method of sharing virtual resources in cooperation with the resource prediction apparatus 120 performed by the cloud server 130 according to the embodiment shown in FIG. 10 is the same as that of the resource prediction system 1 according to the embodiment shown in FIG. 1 And includes steps in which the time is thermally processed. Accordingly, the present invention is also applied to a method of sharing virtual resources in cooperation with the resource prediction apparatus 120 performed by the cloud server 130 according to the embodiment shown in FIGS.

In step S1010, the cloud server 130 may create at least one virtual machine from virtual resources based on the predetermined service level agreement.

In step S1020, the cloud server 130 may process the request from the plurality of user terminals 110 through the generated virtual machine.

In step S1030, the cloud server 130 may dynamically allocate a virtual resource based on the resource prediction value received from the resource prediction apparatus 120. [

In the above description, steps S1010 to S1030 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present invention. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

A method for predicting a resource requirement for a virtual resource in the resource predicting apparatus illustrated in FIGS. 1 to 10 and a method for sharing a virtual resource in cooperation with a resource predicting apparatus in a cloud server may be performed by a computer program Or in the form of a recording medium including instructions executable by a computer. In addition, a method of predicting a resource requirement for a virtual resource in the resource predicting apparatus illustrated in FIGS. 1 to 10 and a method of sharing a virtual resource in cooperation with a resource predicting apparatus in a cloud server are stored in a medium But may also be implemented in the form of a computer program. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. The computer-readable medium may also include computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

110: User terminal
120: resource prediction device
130: Cloud server
210: Penalty Cost Derivative Department
220: resource predicting unit
230: Assignment request unit
710:
720:
730:

Claims (18)

1. An apparatus for predicting a resource requirement for a virtual resource,
A first service penalty cost is derived based on a response time in which a request from a plurality of user terminals is processed by at least one virtual machine created in the cloud server, A penalty cost derivation unit for deriving a second service penalty cost based on a resource sharing amount including a CPU sharing amount and a memory sharing amount;
A resource predicting unit for comparing the first service penalty cost and the second service penalty cost to derive a required resource predicted value; And
An allocation request unit for requesting dynamic allocation of the virtual resource to the cloud server based on the resource prediction value;
The resource prediction device comprising:
The method according to claim 1,
Wherein the resource prediction value includes a first direction value corresponding to the CPU share amount and a second direction value corresponding to the memory share amount.
The method according to claim 1,
Wherein the penalty cost derivation unit derives the first service penalty cost based on whether a response time to a request processed by a virtual machine corresponding to each of the plurality of user terminals exceeds a reference value.
The method of claim 3,
Wherein the penalty cost derivation unit derives the first service penalty cost by applying different weights according to a virtual machine that processes the request.
3. The method of claim 2,
Wherein the resource sharing amount further includes a number of the virtual machines,
Wherein the resource prediction value further includes a third direction value corresponding to the number of the virtual machines.
6. The method of claim 5,
Wherein the resource predicting unit applies the CPU sharing amount, the memory sharing amount, and the number of virtual machines as a component of a three-dimensional array,
And derives the first direction value, the second direction value, and the third direction value from the first service penalty cost and the second service penalty cost using the three-dimensional array.
6. The method of claim 5,
Wherein the resource predicting unit adjusts the second direction value to adjust the first direction value when the memory sharing amount reaches a limit.
8. The method of claim 7,
The resource predicting unit adjusts the third direction value when the CPU sharing amount reaches the limit by adjusting the first direction value,
Wherein the allocation request unit requests the cloud server to additionally generate a virtual machine according to the adjusted third direction value.
6. The method of claim 5,
Wherein the penalty cost derivation unit derives an added infrastructure cost based on at least one of the number and performance of the virtual machines,
And adds the infrastructure cost to the second service penalty cost.
10. The method of claim 9,
Wherein the infrastructure cost is changed according to the third direction value.
11. The method of claim 10,
Wherein the penalty cost derivation unit derives a delay cost based on a delay time generated according to the additional generation of the virtual machine when a virtual machine is additionally generated according to the third direction value,
And adds said delay cost to said second service penalty cost.
1. A cloud server for sharing virtual resources in cooperation with a resource prediction device,
A virtual machine creation unit for creating at least one virtual machine from virtual resources based on a predetermined service level agreement;
A processing unit for processing a request from a plurality of user terminals through the generated virtual machine; And
An allocation unit for dynamically allocating the virtual resource based on the resource prediction value received from the resource prediction apparatus,
Lt; / RTI >
Wherein the resource prediction value includes a first service penalty cost derived based on a response time in which a request from the plurality of user terminals is processed by the virtual machine and a CPU sharing amount provided through a virtual resource of the cloud server And a second service penalty cost derived based on a resource share amount including a memory share amount.
13. The method of claim 12,
Wherein the resource prediction value includes a first direction value corresponding to the CPU share and a second direction value corresponding to the memory share.
13. The method of claim 12,
Wherein the first service penalty cost is derived by weighting according to the virtual machine processing the request.
13. The method of claim 12,
Wherein the resource sharing amount further includes a number of the virtual machines,
Wherein the resource prediction value further comprises a third direction value corresponding to the number of virtual machines.
16. The method of claim 15,
Wherein the third direction value is adjusted based on at least one of adjusting the second direction value to reach a limit for the amount of memory sharing and adjusting a first direction value to reach a limit for the CPU share amount A resource estimating unit,
Wherein the allocator further generates a virtual machine in accordance with the adjusted third direction value.
17. The method of claim 16,
Wherein the allocator dynamically allocates the virtual resources further considering the added infrastructure cost based on at least one of the number and performance of the virtual machines.
18. The method of claim 17,
The allocating unit dynamically allocates the virtual resource considering the delay cost added based on the delay time generated according to the additional generation of the virtual machine when the number of virtual machines is additionally generated according to the third direction value Cloud server.

Images