KR20220003803A - Real-Time Dynamic Resource Allocation Method in a Hybrid Cloud-Based IoT Environment - Google Patents

Abstract

The present embodiments provide a dynamic resource allocation method of a hybrid cloud which receives data traffic and checks a state condition of a hybrid cloud according to the number of activated virtual machines and the number of tasks, thereby obtaining both accuracy and elasticity through a method of allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud.

Description

{Real-Time Dynamic Resource Allocation Method in a Hybrid Cloud-Based IoT Environment}

The technical field to which the present invention pertains relates to a real-time dynamic resource allocation method in a hybrid cloud-based IoT environment.

The content described in this section merely provides background information for the present embodiment and does not constitute the prior art.

A hybrid cloud is a combination of a public cloud and a private cloud environment. A hybrid cloud can give you a lot of control over security, data privacy, and compliance through a private cloud.

The method of allocating computing resources such as virtual machines (VMs) in cloud computing must consider both Accuracy and Elasticity aspects. It is necessary to be able to accurately allocate a virtual machine when a task occurs, and to minimize the time allocated in a dynamic environment. However, there is no resource allocation method that satisfies both accuracy and elasticity in a hybrid cloud environment.

Korean Patent Registration No. 10-1720292 (2017.03.21)

Embodiments of the present invention receive data traffic, check the state condition of the hybrid cloud according to the number of activated virtual machines and the number of tasks, and allocate the minimized number of virtual machines according to data traffic to the hybrid cloud. Through this, the main purpose is to secure both accuracy and elasticity.

Other objects not specified in the present invention may be additionally considered within the scope that can be easily inferred from the following detailed description and effects thereof.

According to an aspect of this embodiment, in a method for allocating a dynamic resource of a hybrid cloud in which a private cloud and a public cloud are combined, receiving data traffic, and the number of activated virtual machines and the number of tasks in the hybrid cloud There is provided a dynamic resource allocation method for performing operations including allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud by checking a state condition.

The state condition may be divided into (i) an expected excess state, (ii) an excess state, (iii) a steady state, (iv) a shortage state, and (v) an expected shortage state.

The step of allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud includes, if the state condition is the expected excess state or the expected insufficient state, assigning the hybrid cloud to the Erlang-C formula The number of minimized virtual machines according to the data traffic can be predicted by modeling as

Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud may include allocating the minimized number of virtual machines to the hybrid cloud after a preset waiting time if the state condition exceeds the expected state. and, if the state condition is the expected shortage state, the minimized number of virtual machines may be immediately allocated to the hybrid cloud.

Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud may include, if the state condition is the excess state or the insufficient state, until the state condition becomes the steady state (Continuous- Time Markov Chains) model can be applied to calculate the minimized number of virtual machines according to the data traffic.

As described above, according to the embodiments of the present invention, the number of minimized virtual machines according to data traffic is received by receiving data traffic and checking the state condition of the hybrid cloud according to the number of activated virtual machines and the number of tasks. By allocating to the hybrid cloud, there is an effect of securing both accuracy and elasticity.

Even if it is an effect not explicitly mentioned herein, the effects described in the following specification expected by the technical features of the present invention and their potential effects are treated as if they were described in the specification of the present invention.

1 is a diagram illustrating a hybrid cloud.
2 is a block diagram illustrating an apparatus for dynamic resource allocation of a hybrid cloud according to an embodiment of the present invention.
3 is a block diagram illustrating a dynamic resource allocation method of a hybrid cloud according to another embodiment of the present invention.
4 to 7 show simulation results performed according to embodiments of the present invention.

Hereinafter, in the description of the present invention, if it is determined that the subject matter of the present invention may be unnecessarily obscure as it is obvious to those skilled in the art with respect to related known functions, the detailed description thereof will be omitted, and some embodiments of the present invention will be described. It will be described in detail with reference to exemplary drawings.

1 is a diagram illustrating a hybrid cloud.

A hybrid cloud is a combination of public cloud and private cloud environments. Data traffic is passed from the Internet of Things (IoT) device to the public cloud through the private cloud. In a hybrid cloud environment, the private cloud acts as a broker that proactively receives data and assigns tasks to multiple clouds.

In cloud computing, the method of allocating computing resources such as virtual machines must consider accuracy. In cloud computing, the Continuous-Time Markov Chains (CTMC) model enables or disables one virtual machine at a time.

The CTMC model uses probabilities to model how the state of an object changes over time. The Markov chain describes how to connect the different states of objects over time, and the Markov matrix serves as a medium that connects them. If the time parameter is continuous time, it corresponds to a Markov process.

In cloud computing, the method of allocating computing resources such as virtual machines must consider elasticity. Although the resource allocation method using the CTMC model is accurate, it is necessary to supplement the elasticity aspect.

The dynamic resource allocation apparatus according to this embodiment models the hybrid cloud environment as an Erlang-C formula or M/M/c queue, and applies a KKT (Karush-Kuhn-Tucker) model to elastically allocate resources. make it possible Even if the hybrid cloud environment is changed to use an external broker, since there is no difference in the operation algorithm, the dynamic resource allocation apparatus according to the present embodiment can be applied as it is.

2 is a block diagram illustrating an apparatus for dynamic resource allocation of a hybrid cloud according to an embodiment of the present invention.

The dynamic resource allocation apparatus 110 includes at least one processor 120 , a computer readable storage medium 130 , and a communication bus 170 .

The processor 120 may control to operate as the dynamic resource allocation apparatus 110 . For example, the processor 120 may execute one or more programs stored in the computer-readable storage medium 130 . The one or more programs may include one or more computer-executable instructions, which when executed by the processor 120 configure the dynamic resource allocation apparatus 110 to perform operations in accordance with the exemplary embodiment. can be

Computer-readable storage medium 130 is configured to store computer-executable instructions or program code, program data, and/or other suitable form of information. The program 140 stored in the computer-readable storage medium 130 includes a set of instructions executable by the processor 120 . In one embodiment, computer-readable storage medium 130 includes memory (volatile memory, such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, It may be a flash memory device, other form of storage medium that is accessed by the breathing interval detection device 110 and can store desired information, or a suitable combination thereof.

Communication bus 170 interconnects various other components of dynamic resource allocation device 110 including processor 120 and computer readable storage medium 140 .

The dynamic resource allocation device 110 may also include one or more input/output interfaces 150 and one or more communication interfaces 160 that provide interfaces for one or more input/output devices 24 . The input/output interface 150 and the communication interface 160 are connected to the communication bus 170 . The input/output device (not shown) may be connected to other components of the dynamic resource allocation device 110 through the input/output interface 150 .

The dynamic resource allocation device 110 models the hybrid cloud environment as an Erlang-C formula or M/M/c queue and applies a KKT (Karush-Kuhn-Tucker) model to enable elastic resource allocation. do.

The Erlang-C formula is a formula that defines the probability that a customer can receive service when a customer call arrives at the system according to a Poisson distribution. When the system receives a customer's service request, when all service counters are busy, the system stores the customer in a queue and provides the service in turn. The form of the queue may be defined as an M/M/c queue or the like.

In an optimization problem with constraints, the simultaneous inequality constraint satisfies the KKT condition. The KKT (Karush-Kuhn-Tucker) condition is (1) that the derivative of _{all independent variables x 1} , x ₂ , ... , x _{N is 0.} (2) The product of all Lagrange multipliers λ ₁ , ... ,λ _M and the constraint inequality (differential value with respect to λ) is zero. (3) Lagrange multipliers must not be negative.

In order to solve the problem that the KKT model cannot be applied to the Erlang-C formula, the dynamic resource allocation device 110 applies the KKT model through approximate prediction for the Erlang-C formula.

Cloud computing cannot avoid network delays as data is exchanged between virtual machines. Equipment delay (equipment delay, T _equip ) may be expressed as a constant value. The propagation delay (T _prop ) may be derived by dividing the distance between the sender and the virtual machine. The propagation delay T _prop is independent of the data size. The transmission delay (T _trans ) is affected by the data size.

L is the data size (in bits) and R is the average terminal bandwidth (in bits per second, bps). The terminating network delay is expressed as _{T equip} + T _prop + T _{trans .}

In the hybrid cloud model, there are a plurality of IoT devices that request task processing. It may be assumed that the arrival of _{task X k} from the k-th IoT device follows a Poisson distribution.

λ is the average arrival rate, and n is the total number of IoT devices. In the j-th cloud, the service speeds of virtual machines are the same and follow an exponential distribution with _{mean μ j .} Each cloud can be treated as an M/M/c queue. Referring to the cloud service speed μ _j , index j=0 means a private cloud, and if index j is a natural number other than 0, it means the j-th public cloud.

The processed data size may be varied so, represented by L _O average data of the input data from the IoT device _I L, the average output data is multiplied by the coefficient data traffic (data traffic coefficient, DTC).

Let the data traffic coefficient from the private cloud to the j-th cloud be α _j and the data traffic coefficient in the opposite direction is β _j .

The installation cost of a hybrid cloud is double the cost for a private cloud and the cost for a public cloud. The cost for the public cloud is expressed as Equation 4, and the cost for the private cloud is expressed as Equation 5.

는 렌탈 비용과 고정 소비 전력의 합이고,

는 동적 소비 전력이고, N_j는 j 번째 클라우드에서 가상 머신의 개수이다.

is the sum of rental cost and fixed power consumption,

is the dynamic power consumption, and N _j is the number of virtual machines in the j-th cloud.

는 가상 머신 별 구매 비용이다. 구매 비용은 네트워크 장비 비용, 가상 머신 소프트웨어 라이센스, 애플리케이션 소프트웨어 라이선스를 포함한다.

는 동적 소비 전력 총합이다.N ₀ is the number of virtual machines in the private cloud,

is the purchase cost per virtual machine. The purchase cost includes the cost of network equipment, virtual machine software licenses, and application software licenses.

is the sum of dynamic power consumption.

과

으로 치환한다.In Equation 6, parentheses

class

replace with

The combination of pre-processing and scheduling in the cloud model is called groundwork. The service time of the preparatory work can be assumed to be exponentially distributed with the _{parameter μ G .} Any virtual machine in the private cloud can perform the necessary tasks and complete the tasks assigned to it by performing the remaining necessary tasks. It is assumed that the service time of the preparation work is relatively small compared to the service time required to complete the remaining tasks.

Since the incoming input data traffic follows a Poisson distribution, the preparation of the private cloud can be modeled as an _{M/M/N 0 queue model.} The delay τ _G of the preparation work is expressed as in Equation (8).

C( ) is defined by the Erlang-C formula and can be expressed as Equation (7).

The virtual machine usage ρ _G must be less than 1, and the queue grows to infinity.

When the preparatory work is completed, the private cloud transmits the pre-processing data to the public cloud as needed. The transmission delay caused by transmitting data from the private cloud to the j-th public cloud is expressed as Equation 11.

The delay time is expressed as Equation (12).

After network processing, the public cloud handles the remaining tasks.

(Theorem 1) The arrival of data to the j-th public cloud follows a Poisson distribution with an _{average velocity ω j λ.} Each public cloud can be modeled with an M/M/N _{j queue model.} In the j-th public cloud, the execution time is expressed as Equation 13.

C() is defined by the Erlang-C formula and is expressed as Equation 14.

When the j-th public cloud completes the remaining tasks, the j-th public cloud sends the output data back to the private cloud. The delay time considering the transmission delay is expressed as Equation (15).

The time required to process the remaining task using the j-th public cloud is expressed as Equation 16.

τ _N = τ _0j + τ _j0 .

Preparatory work is also required when using a private cloud.

C() is defined by the Erlang-C formula and is expressed as Equation 18.

In the private cloud, τ _N = 0.

를 찾는 것이다.The least cost hybrid cloud computing approach must complete tasks within the range ^{of delay threshold D*.} The goal of task P1 is to minimize the number of virtual machines that satisfy the delay threshold.

is to find

The vector N = (N ₀ , N ₁ , ... , N _M ).

,

,

상한을 설정하면 수학식 20, 21, 22와 같이 표현된다.In Equations 8, 13, and 17, since the Erlang-C formula is not differentiable, the KKT condition cannot be applied to the problem P1. The dynamic resource allocation apparatus according to the present embodiment sets an upper limit in the Erlang-C formula. About Equations 9, 14 and 18

,

,

When the upper limit is set, it is expressed as Equations 20, 21, and 22.

Differentiating Equations 20, 21 and 22 is expressed as Equations 23, 24 and 25.

The upper limit of the delay obtained by applying the upper limit of the Erlang-C formula to the delay time can be expressed as Equations 26, 27, and 28.

The vector g is expressed as in Equation 29.

을 설정할 수 있다. g는 미분 가능하므로, 해결과제 P1을 해결과제 P2로 치환할 수 있다. 해결과제 P2에 의해 3 개의 정리가 성립될 수 있다.Inequalities for all j through equations 29, 30 and 31

can be set. Since g is differentiable, the solution P1 can be substituted for the solution P2. Three theorems can be established by the solution task P2.

(Theorem 2) The solution found while solving the problem P2 is sufficient to satisfy the constraint of the problem P1.

(Theorem 3) The inequality constraints of task P2 are all convex.

를 참조하면

는 수학식 32 및 수학식 33과 같이 표현되고, 분수를 분해하면 수학식 34와 같이 표현된다.

refer to

is expressed as in Equations 32 and 33, and when the fraction is decomposed, it is expressed as Equation 34.

는 해결과제 P2의 최적화된 전역 해(global optimal solutions)이다. (Theorem 4) Using KKT to find the optimized value,

is the global optimal solutions of the problem P2.

The KKT condition for the solution task P2 is expressed as Equations 35, 36, and 37.

로 정의된 KKT의 곱셈 벡터이다.* represents the solution, π is

It is a multiplication vector of KKT defined as .

The validity of the solution is expressed as in Equation 38.

In order to satisfy the condition of Equation 37, Equation 42 must be satisfied.

The service time of the preparatory work is relatively small compared to the service time required to complete the remaining tasks.

는 임의의 작은 수

보다 작다.

is any small number

smaller than

Through this process, the solution is expressed as Equation 46.

An automatic scaling method can be applied in cloud computing.

m _j is the number of virtual machines activated in the j-th cloud.

k _j is the number of tasks in the j-th cloud.

(m _j , k _j ) is the state of the j-th cloud.

[a(m _j ), b(m _j )] is an integer that determines the state of the j-th cloud. b(m _j ) > a(m _j ) >= m-1, b(m _j ) >= a((m+1) _j ), a(1 _j ) < a(2 _j ) < a(3 _{j )} ) < ... , b(1 _j ) < b(2 _j ) < b(3 _j ) < ...

Time-series analysis is not suitable for periodic workloads. For example, IoT devices access the cloud much more during the day than during the night. In order to be able to consider these periodic workloads, it further includes a likely-over-provisioned state and a likely-under-provisioned state using the _{indicators o(m j} ) and u(m _{j ).} Set the state condition.

및

으로 설정되고,

및

는 조절 가능한 임의의 양의 실수이다. o(m_j)와 u(m_j)는 주기적 워크로드 상황에서 활성화된 IoT 장치의 개수를 예측하는 데 적합하다. The indicators o(m _j ) and u(m _j ) are

and

is set to

and

is any adjustable real number. o(m _j ) and u(m _j ) are suitable for predicting the number of active IoT devices in a periodic workload situation.

3 is a block diagram illustrating a dynamic resource allocation method of a hybrid cloud according to another embodiment of the present invention. The hybrid cloud dynamic resource allocation method may be performed by a hybrid cloud dynamic resource allocation apparatus.

In step S10, the dynamic resource allocation method receives data traffic.

In step S20, the dynamic resource allocation method determines whether the state condition of the cloud corresponds to a normal state. If the state condition does not correspond to the normal state, the dynamic resource allocation method checks the number of virtual machines and the number of tasks activated in the hybrid cloud (S30).

In step S40, the dynamic resource allocation method determines whether the state condition of the cloud corresponds to a likely-over-provisioned state. If the state condition corresponds to the expected excess state, the dynamic resource allocation method allocates the optimized number after the waiting time by modeling the Erlang-C formula, applying the upper limit, applying the differential to the upper limit, and applying the KKT condition (S50) .

In step S60, the dynamic resource allocation method determines whether the state condition of the cloud corresponds to a likely-under-provisioned state. If the state condition corresponds to the expected shortage state, the dynamic resource allocation method immediately allocates the optimized number by modeling the Erlang-C formula, applying the upper limit, applying the differential to the upper limit, and applying the KKT condition (S70).

In step S80, the dynamic resource allocation method determines whether the state condition of the cloud corresponds to an over-provisioned state or an under-provisioned state. If the state condition corresponds to the excess state or the insufficient state, the dynamic resource allocation method allocates an optimized number by applying the CTMC model until the state condition becomes a normal state (S90).

Table 1 shows the dynamic resource allocation method using pseudo codes.

4 to 7 show simulation results performed according to embodiments of the present invention.

The elasticity measurement of resource allocation of cloud computing uses Equation 51.

As shown in FIGS. 4 to 7 , it can be easily understood that the dynamic resource allocation model (MARS) according to the present embodiment has a superior effect in terms of elasticity than the CTMC model.

The dynamic resource allocation apparatus may be implemented in a logic circuit by hardware, firmware, software, or a combination thereof, or may be implemented using a general-purpose or special-purpose computer. The device may be implemented using a hardwired device, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or the like. In addition, the device may be implemented as a system on chip (SoC) including one or more processors and controllers.

The dynamic resource allocation apparatus may be mounted on a computing device or server provided with hardware elements in the form of software, hardware, or a combination thereof. A computing device or server is all or part of a communication device such as a communication modem for performing communication with various devices or wired/wireless communication networks, a memory for storing data for executing a program, and a microprocessor for executing operations and commands by executing the program It can mean a variety of devices, including

Although it is described that each process is sequentially executed in FIG. 3, this is only illustratively described, and those skilled in the art change the order described in FIG. Alternatively, various modifications and variations may be applied by executing one or more processes in parallel or adding other processes.

The operations according to the present embodiments may be implemented in the form of program instructions that can be performed through various computer means and recorded in a computer-readable medium. Computer-readable medium represents any medium that participates in providing instructions to a processor for execution. Computer-readable media may include program instructions, data files, data structures, or a combination thereof. For example, there may be a magnetic medium, an optical recording medium, a memory, and the like. A computer program may be distributed over a networked computer system so that computer readable code is stored and executed in a distributed manner. Functional programs, codes, and code segments for implementing the present embodiment may be easily inferred by programmers in the art to which this embodiment belongs.

The present embodiments are for explaining the technical idea of the present embodiment, and the scope of the technical idea of the present embodiment is not limited by these embodiments. The protection scope of this embodiment should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present embodiment.

Claims (5)

In the dynamic resource allocation method of a hybrid cloud in which a private cloud and a public cloud are combined,
receiving data traffic; and
Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud by checking a state condition of the hybrid cloud according to the number of activated virtual machines and the number of tasks
A dynamic resource allocation method for performing operations including According to claim 1,
The state condition is (i) an expected excess state, (ii) an excess state, (iii) a steady state, (iv) a shortage state, and (v) an expected shortage state. 3. The method of claim 2,
Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud comprises:
If the state condition is the expected excess state or the expected insufficient state, the hybrid cloud is modeled with an Erlang-C formula, an upper bound is applied to the Erlang-C formula, and a derivative is applied to the upper bound, and A dynamic resource allocation method of a hybrid cloud, characterized in that the number of minimized virtual machines according to the data traffic is predicted by applying a Karush-Kuhn-Tucker (KKT) condition. 3. The method of claim 2,
Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud comprises:
If the state condition is the expected excess state, allocating the minimized number of virtual machines to the hybrid cloud after a preset waiting time,
When the state condition is the expected shortage state, the dynamic resource allocation method of the hybrid cloud, characterized in that immediately allocating the minimized number of virtual machines to the hybrid cloud. 3. The method of claim 2,
Allocating the minimized number of virtual machines according to the data traffic to the hybrid cloud includes:
If the state condition is the excess state or the insufficient state, the number of minimized virtual machines according to the data traffic is calculated by applying a Continuous-Time Markov Chains (CTMC) model until the state condition becomes the steady state. Hybrid cloud dynamic resource allocation method, characterized in that.

Images