TW201525717A - A method for adapting resource of the cloud service - Google Patents

Abstract

This invention discloses a method for adapting resource of the cloud service is disclosed to solve the estimation problem of the machine cost and the integral performance of the cloud operation environment. The method is executed by a computer system. The method comprises setting a limit cost, a service request, a level quantity, and a level cost, an average time, a request allotment, an execution per time and a purchase per time of each of a plurality service level, the quantity of the level cost, the average time, the request allotment, the execution per time and the purchase per time are equal to the level quantity, adjusting the request allotment and the execution per time of each request level, changing the execution per time of each virtual machine to an adapting execution and adjusting the request allotment and adapting execution of each service level. Thus, it can actually solve the said problem.

Description

Resource adaptation method for cloud services

The present invention relates to a resource adaptation method; in particular, to a resource adaptation method for a cloud service.

Cloud computing is a technology that delivers resources to users over the Internet in a service-based manner. It can be divided into "IaS for Infrastructure (IaaS)" and "Passion for Service (PaaS)". "Software as a Service (SaaS)". The Web Service technology provides services through Web communication protocols (such as HTTP, HTTPS, etc.) and Open Standards (such as XML, XML Schema, etc.), including: simple objects. Three basic components, such as the Simple Object Access Protocol (SOAP), the Web Service Description Language (WSDL), and the Universal Description Discovery and Integration (UDDI). Figure shows). Compared with traditional intermediary software technologies (such as DCOM, CORBA, etc.), network service technology is more versatile and integrated.

Among them, Virtualization Technique is one of the key technologies of cloud computing. The virtual machine itself is a complete system platform with operating system or application service environment, and the resources allocated by the virtual machine are reusable and adjustable. Sexuality to reconfigure resources or virtual machines as demand changes. However, the resource adaptation method of the cloud service is to deploy the network service directly on the physical machine, which is easily affected by many external factors, and causes all network services on the physical machine to be shut down, indirectly causing physical machine resources. wave Fees and the instability of system or application services.

In view of this, it is necessary to propose a resource adaptation method for cloud services. Even if one or more virtual machines cannot be operated on the same physical machine, the virtual machine rental cost and overall performance can be considered, and the change according to the total demand of virtual machines can be considered. To adjust the number of virtual machines of different levels without affecting the operation of other virtual machines and improving their practicality.

The main purpose of the present invention is to provide a resource adaptation method for a cloud service, which considers virtual machine rental cost and overall performance under variable resources, and adjusts the number of virtual machines of different levels according to changes in the total demand of virtual machines. .

The invention provides a resource adaptation method for a cloud service, which is executed by a computer system, and includes: setting a limit cost, a service demand quantity, a level quantity, a plurality of service level level costs, an average time, a demand allocation quantity, and a single The quantity of the execution and the quantity of the single purchase, the quantity of the level, the average time, the quantity of the demand, the quantity of the single execution, and the quantity of the single purchase are the quantity of the level; adjusting the demand allocation of each service level and A single execution amount until the following conditions are met: (1) The sum of the demand distributions is equal to the service demand, and (2) the single execution quantity of each service level is not greater than the single purchase quantity, (3) Calculate the demand distribution amount of each service level and the quotient of the single execution quantity as the batch quantity of each service level, calculate the product of the batch quantity and the average time of each service level as the execution time of each service level, and make each execution time The value of the largest one is the smallest, the largest of each execution time takes the minimum value as a minimum execution time, and (4) the product of the single execution quantity of each service level and the level cost. If the cost is greater than the limit cost, the product of the single execution amount of each virtual machine and the level cost is set to be a first execution cost; and the single execution amount of each virtual machine is changed to an adjustment execution amount, and the requirements of each service level are adjusted. The amount of allocation and the amount of execution are adjusted until the following conditions are met: (1) the sum of the demand distributions is equal to the service demand, and (2) the adjustment execution amount of each virtual machine is not greater than the single purchase amount, ( 3) Calculate the demand distribution amount of each service level and the quotient of the adjustment execution amount as the batch number of each virtual machine The quantity is calculated by multiplying the number of batches of each service level by the average time as the execution time of each service level, and the maximum value of each execution time is not greater than the minimum execution time, and (4) calculating the adjustment execution amount of each service level and The sum of the product of the rank cost, the minimum value of the sum of the products is an optimal cost combination, and the optimal cost combination is not greater than the first execution cost.

Preferably, the method further comprises: adjusting the demand allocation amount of each service level and adjusting the execution amount until the following conditions are met simultaneously: (1) the sum of the plurality of demand allocation amounts is equal to the service demand quantity, and (2) each virtual quantity The adjustment execution amount of the machine is not greater than the single purchase amount, (3) calculating the demand distribution amount of each service level and the quotient of the adjustment execution amount as the batch quantity of each virtual machine, and calculating the batch quantity and the average time of each service level. The product is used as the execution time of each service level, and the maximum value of each execution time is not greater than the minimum execution time. (4) The product sum of the adjustment execution amount of each service level and the level cost is calculated, and the minimum value of each product sum is one. The optimal cost combination, the optimal cost combination is not greater than the limited cost, (5) calculating the demand allocation amount of each service level and the modulus of the adjustment execution amount, and calculating the product of the modulus and the batch number of each service level as one The first provisional value is obtained by summing the number of single purchases of the batches and the first provisional value at each service level, and multiplying the adjustment execution amount of each service level as a second provisional value, and calculating each Product of the second temporary value and the average level of service and the time taken for each product and for the minimum time to an optimum performance.

Preferably, the minimum execution time is calculated as shown in the following equations (1) and (2): Wherein, T _execute for the minimum execution time, l is the number of each class of service, L is the number of all the service levels, the total computation requirement for the service _{q, r 1, ..., r l} , ..., r L for The total amount of calculation of the service demand is the amount of allocation at each service level, and u ₁ , ..., u _l , ..., u _L are the amount of calculations actually performed each time the service level is performed. For the number of batches for each service level, For the execution time of each service level, n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level, and t ₁ , ..., t _l , ..., t _L are the service levels of each service level. The average time, c _l is the service cost of each level of virtual machine, C is the limit cost, and C' is the first execution cost.

Preferably, the calculation of the optimal cost combination is as follows: Wherein, T _execute for the minimum execution time, l is the number of each class of service, L is the number of all the service levels, the total computation requirement for the service _{q, r 1, ..., r l} , ..., r L for The total amount of operation demand for service is allocated at each service level, and u ₁ ', ..., u _l ', ..., u _L ' is the amount of adjustment for each service level. For the number of batches for each service level, For the execution time of each service level, n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level, and t ₁ , ..., t _l , ..., t _L are the service levels of each service level. The average time, c _l is the service cost of each level of virtual machine, C' is the first execution cost, and C' is the optimal cost combination.

Preferably, the optimal performance time is calculated as follows: Where T _{total is} the optimal performance time, _{Texem is} the minimum execution time, l is the number of each service level, L is the number of all service levels, q is the total operation amount of the service demand, r ₁ , ..., r _l , ..., r _{L are} the distribution amount of the total operation amount of the service demand at each service level, and u ₁ ', ..., u _l ', ..., u _L ' are the adjustment execution amounts of the service levels, For the number of batches for each service level, For the execution time of each service level, n _l is the single execution quantity currently purchased by each service level, and t ₁ , ..., t _l , ..., t _L are the average time of each service level, For the first provisional value, For the second provisional value, C' is the first execution cost, and C' is the optimal cost combination.

〔this invention〕

1‧‧‧Application layer

2‧‧‧Load Balancing Intermediary

3‧‧‧Web service layer

4‧‧‧Virtual Machine Layer

S1‧‧‧ parameter setting procedure

S2‧‧‧First operation steps

S3‧‧‧Second operation steps

S4‧‧‧ third operation steps

Figure 1 is a system architecture diagram of a conventional network service.

FIG. 2 is a system architecture diagram of an embodiment of a resource adaptation method for a cloud service according to the present invention.

FIG. 3 is a flow chart showing the operation of an embodiment of the resource adaptation method of the cloud service of the present invention.

Figure 4 is a schematic diagram showing the minimum execution time of an embodiment of the resource adaptation method of the cloud service of the present invention.

FIG. 5 is a schematic diagram of an optimal cost combination of an embodiment of a resource adaptation method for a cloud service according to the present invention.

FIG. 6 is a schematic diagram showing the optimal performance time of an embodiment of the resource adaptation method of the cloud service of the present invention.

The above and other objects, features and advantages of the present invention will become more <RTIgt; Cost (C) refers to the total cost of service (unit: yuan) used to limit a service request (such as server, virtual machine, etc.), which is in the technical field of the present invention. Usually the knowledge person can understand.

The "service demand amount (q)" as used throughout the present invention refers to the total amount of calculation (unit: number of times) of a service demand, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

The "number of levels (L)" described throughout the present invention refers to the number of resource levels of a virtual machine, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

The "level cost (c)" as described throughout the present invention refers to the service cost (unit: unit) of each level of virtual machine, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

The "average time (t)" as described throughout the present invention refers to the average service time (unit: second) used by each level of virtual machine for the service demand, as will be understood by those of ordinary skill in the art to which the present invention pertains.

The "demand allocation amount (r)" as described throughout the present invention refers to the amount of allocation of the service demand (q) to each level of virtual machine, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

The "single execution amount (u)" as used throughout the present invention refers to the amount of calculation (unit: number of times) actually executed by each level of virtual machine at a time, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

The "single purchase amount (n)" as described throughout the present invention refers to a single execution amount (unit: number of times) currently purchased by each level of virtual machine, which can be understood by those having ordinary knowledge in the technical field to which the present invention pertains.

Please refer to FIG. 2 , which is a system architecture diagram of an embodiment of a resource adaptation method for a cloud service according to the present invention. Among them, a cloud system can be planned by Software/Hardware into a Client Application Layer, a Service Load Balance Middleware Layer 2, and a Web Service Layer. 3) and a virtual machine layer (Virtual Machine Layer) 4, as an execution architecture of the resource adaptation method of the cloud service of the present invention, wherein the network layering system providing the cloud service is familiar with the skilled person by using soft/hardware technology. It can be understood that it will not be described here. In this embodiment, the application layer 1 is developed for the developer to meet the needs of the user, for example, by combining the Rich Internet Application (RIA) concept, allowing the user to directly browse. The Web Browser can use the developed Cloud Service; the load balancing agent layer 2 can consider the virtual machine rental cost and the overall performance under the variable resources, and according to the change of the total demand of the virtual machine. The adjustment of the number of different levels of virtual machines; the network service layer 3 can be configured by the service provider to build a plurality of physical machines, and the provided network services can be uploaded through the load balancing intermediation layer 2 and deployed to the virtual machine layer. 4; the virtual machine layer 4 can use virtualized software (such as: Virtualbox, VMware, Xen, etc.), the physical machine resources (such as: data processing or storage resources, etc.) are divided into several virtual machines, and become an operating system (Operation System) or a complete system platform of an application service environment, the resources allocated by the virtual machine are reusable and tunable to Changing business needs and re-allocation of resources.

Referring to FIG. 3, it is an operational flowchart of an embodiment of a resource adaptation method for a cloud service according to the present invention. The resource adaptation method of the cloud service may be performed by a computer system, including a parameter setting step S1 and a first An operation step S2, a second operation step S3, and a third operation step S4 are respectively described later, please refer to FIG. 2 together.

The parameter setting step S1 sets a limit cost (C), a service demand amount (q), a level quantity (L), a level cost (c) of several service levels, an average time (t), and a demand allocation amount. (r), single execution amount (u) and single purchase amount (n), the level cost (c), average time (t), demand allocation amount (r), single execution amount (u), single time The quantity of purchase quantity (n) is the quantity of this level (L). In detail, the cost limit (C) is used to limit the service time of a service request (such as a server, a virtual machine, etc.) for a service request; the service demand q is a service requirement (such as : the total amount of calculation of the data search, etc.; the number of levels (L) is the number of resource levels of the virtual machine; the cost of the level (c) is the service cost of each level of virtual machines (unit: yuan); the average time (t The average service time (in seconds) used by each level of virtual machines for the service demand; the demand allocation amount (r) is the allocation amount of the service demand (q) at each level of the virtual machine; the single execution amount (u) is the amount of computation actually performed by each level of virtual machine each time; the single purchase amount (n) is a single execution amount currently purchased by each level of virtual machine. Thereafter, the first operational step S2 is performed.

Please refer to FIG. 3 again, wherein the first operation step S2 adjusts the demand allocation amount (r) and the single execution amount (u) of each service level until the following conditions are met simultaneously: [1] The sum of several demand allocations (r) is equal to the service demand (q); [2] the single execution quantity (u) of each service level is not greater than the single purchase quantity (n); [3] calculate each service The quotient of the demand distribution amount (r) and the single execution amount (u) is the number of batches of each service level, and the product of the batch number of each service level and the average time (t) is calculated as the execution time of each service level. , the largest value of each execution time is the smallest, and the largest one of each execution time takes the minimum value as a minimum execution time; [4] the product of the single execution quantity (u) of each service level and the level cost (c) And not more than the limit cost, setting the product sum of the single execution amount (u) of each virtual machine and the level cost (c) as a first execution cost. In detail, the calculation method of the minimum execution time and the first execution cost are respectively as shown in the following formulas (1) and (2): Where T _{execute is} the minimum execution time (unit: second); l is the number of each service level; L is the number of all service levels; q is the total operation amount of the service demand; r ₁ ,..., r _l ,... r _{L is} the allocation amount of the total operation amount of the service demand at each service level; u ₁ , ..., u _l , ..., u _L are the calculation amounts actually executed each time the service level is performed; The number of batches for each service level; The execution time for each service level; n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level; t ₁ , ..., t _l , ..., t _L are the service levels Average time; c _l is the service cost of each level of virtual machine (unit: yuan); C is the limit cost; C' is the first execution cost. Thereafter, the second operational step S3 is performed.

Referring to FIG. 4, which is a schematic diagram of a minimum execution time of an embodiment of a resource adaptation method for a cloud service according to the present invention, wherein a demand allocation amount (r) and a single execution amount (u) of each service level are adjusted, The execution time of the service level 1~4 is the minimum value respectively. , , , The largest is the execution time of service level 1. , you can take the execution time As the minimum execution time, and the first execution cost is not greater than the limit cost.

Please refer to FIG. 3 again, wherein the second operation step S3 is to change the single execution amount (u) of each virtual machine to an adjustment execution amount (u'), and adjust the demand allocation amount of each service level. (r) and adapt the execution amount (u') until the following conditions are met: [1] The sum of the several demand allocations (r) is equal to the service demand (q); [2] for each virtual machine The adjustment execution amount (u') is not greater than the single purchase amount (n); [3] the quotient of the demand distribution amount (r) and the adaptation execution amount (u') for each service level is calculated as the batch number of each virtual machine. Calculating the product of the batch number of each service level and the average time (t) as the execution time of each service level, and the maximum value of each execution time is not greater than the minimum execution time; [4] calculating the adjustment execution amount of each service level The product sum of (u') and the level cost (c), the minimum value of each product sum is an optimal cost combination, and the optimal cost combination is not greater than the first execution cost. In detail, the calculation of the optimal cost combination is as shown in the following equation (3): Where T _{execute is} the minimum execution time (unit: second); l is the number of each service level; L is the number of all service levels; q is the total operation amount of the service demand; r ₁ ,..., r _l ,... r _{L is} the allocation amount of the total calculation amount of the service demand at each service level; u ₁ ', ..., u _l ', ..., u _L ' is the adjustment execution amount of each service level; The number of batches for each service level; The execution time for each service level; n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level: t ₁ , ..., t _l , ..., t _L are the service levels The average time; c _l is the service cost (unit: yuan) of each level of virtual machine; C' is the first execution cost; C" is the optimal cost combination. Thereafter, the third operation step S4 can also be performed.

Please refer to FIG. 5 , which is a schematic diagram of an optimal cost combination of an embodiment of a resource adaptation method for a cloud service according to the present invention, wherein the resources provided by the service level 4 can complete the service requirement faster, but the required The cost is higher. Therefore, after measuring the minimum execution time and the service demand, the service resource of the service level 1 is selected to be added, and the service resource of the service level 2 and 3 is used as the optimal cost combination to use the service level resource with high price. Adjust to a low-cost service level resource to achieve the required service needs.

Please refer to FIG. 3 again, wherein the third operation step S4 adjusts the demand allocation amount (r) and the adjustment execution amount (u') of each service level until the following conditions are met simultaneously: [1] The sum of several demand allocations (r) is equal to the service demand (q); [2] the adaptive execution amount (u') of each virtual machine is not greater than the single purchase amount (n); [3] calculate each service The quotient of the demand distribution amount (r) and the adjustment execution amount (u') is used as the batch number of each virtual machine, and the product of the batch number of each service level and the average time (t) is calculated as the execution time of each service level. The maximum value of each execution time is not greater than the minimum execution time; [4] calculating the product sum of the adjustment execution amount (u') of each service level and the level cost (c), and the minimum value of each product sum is one of the most a combination of cost, the optimal cost combination is not greater than the limit cost; [5] calculating the modulus of demand distribution (r) and adaptation execution amount (u') for each service level, and calculating the modulus and batch of each service level The product of the second quantity is taken as a first provisional value, and the quantity of the batch is taken at each service level (n) And summing the first provisional value, and multiplying the adjustment execution amount (u') of each service level as a second provisional value, and calculating a product sum of the second provisional value of each service level and the average time (t), The minimum of each product sum is an optimum performance time. In detail, the best performance time is calculated as shown in equation (4): Where T _{total is} the optimal performance time; T _{execute is} the minimum execution time (unit: second); l is the number of each service level; L is the number of all service levels; q is the total operation amount of the service demand; r ₁ , ..., r _l , ..., r _{L are} the total amount of operations required for the service at each service level; u ₁ ', ..., u _l ', ..., u _L ' is the implementation of each service level the amount; The number of batches for each service level; The execution time for each service level; n _l is the single execution amount currently purchased by each service level; t ₁ , ..., t _l , ..., t _L are the average time of each service level; For the first provisional value; For the second provisional value; C' is the first execution cost; C" is the optimal cost combination.

Please refer to FIG. 6 , which is a schematic diagram of an optimal performance time of an embodiment of a resource adaptation method for a cloud service according to the present invention, wherein the minimum execution time is a fixed value due to consideration of the purchased service resources. Therefore, the minimum execution time changes only when the service resources that have been purchased are not considered, so the overall execution time after the resource adjustment needs to be calculated as the optimal performance time.

For example, assume a limit cost of C = 15 yuan per hour, service demand q=140, the execution time and price of different levels of virtual machines are shown in Table 1 below.

Among them, since no virtual machine of any level has been purchased at the beginning, the minimum execution time _{Texem is} calculated in the case of meeting the demand quantity and the total cost, as shown in the following formula.

The minimum execution time T _execute = 1.5 and the first execution cost C'=14.793 have been obtained. In the case of meeting the minimum execution time, the cost is not necessarily the minimum value, and the combination may be as shown in Table 2 below.

Then, the minimum execution time T _execute = 1.5 is used to obtain the minimum value, and the cost combination is adjusted to the best, as shown in the following equation.

In the case where the minimum execution time T _execute = 1.5 and the minimum cost C"=14.336 are obtained, the optimal cost combination is adjusted as shown in Table 3 below.

However, when the demand change is 120, considering the quantity that has been purchased, the calculation can be as follows.

The minimum execution time T _execute = 1.5 and C'=14.183 are obtained. If the minimum execution time is met, the cost is not necessarily the minimum value, and may be combined as follows. Therefore, the minimum execution time T _execute = 1.5 is used to obtain the minimum cost, and the cost combination is adjusted to the best.

Under the minimum execution time T _execute = 1.5, the minimum cost C" = 12.2, adjusted to the best combination is as follows.

Table 7 Best Cost Combination Table As shown in Table 7, it is assumed that the virtual machine demand of service level 3 is allocated to 120, the executable amount is 44 each time, and the execution time is 1 second each time, and the total execution time is calculated as shown in the following formula.

Since only 44 requests can be executed at a time, 44x1 is the number that was executed for the first time, and 44 x 2 is the number of the second execution, so it is necessary to add 44 times of the previous execution as the waiting time, 88 The requirements are executed as 1 x (44 x1 + 44 x 2), and the remaining 32 requests are executed for the third time, so the first two wait times (as shown below) need to be added.

The total execution time is 1 x (32 x 3), and the total execution time of 120 requirements is 1 x [(44 x 1+44x2) + (32 x 3)]. The total execution time of all levels is as follows. Show.

The main features of the resource adaptation method embodiment of the cloud service of the present invention are as follows: First, the above-mentioned limit cost (C), service demand (q), number of levels (L), and several services are set. Level cost (c), average time (t), demand allocation (r), single execution amount (u), and single purchase amount (n), the level cost (c), average time (t), the demand allocation amount (r), the single execution amount (u), and the single purchase amount (n) are the number of the level (L). Then, the demand allocation amount (r) and the single execution amount (u) of each service level are adjusted until the following conditions are met: [1] The sum of the several demand allocation amounts (r) is equal to the service demand amount ( q); [2] The single execution amount (u) of each service level is not greater than the single purchase amount (n); [3] calculate the demand allocation amount (r) and the single execution amount (u) for each service level. As the batch quantity of each service level, the business calculates the product of the batch quantity of each service level and the average time (t) as the execution time of each service level, so that the largest value of each execution time is the smallest, and each execution time is The maximum value is the minimum execution time; [4] The product of the single execution amount (u) of each service level and the level cost (c) is not greater than the limit cost, and the single execution amount of each virtual machine is set ( u) The product sum of the level cost (c) is the first execution cost described above. After that, the single execution amount (u) of each virtual machine is changed to the above-described adjustment execution amount (u'), and the demand allocation amount (r) and the adjustment execution amount (u') of each service level are adjusted until the following conditions are simultaneously met. So far: [1] the sum of the several demand allocations (r) is equal to the service demand (q); [2] the adaptive execution amount (u') of each virtual machine is not greater than the single purchase amount (n); [3] Calculate the quotient of the demand allocation amount (r) and the adjustment execution amount (u') for each service level as the batch number of each virtual machine, and calculate the product of the batch number of each service level and the average time (t) as The execution time of each service level, the maximum value of each execution time is not greater than the minimum execution time; [4] calculating the product sum of the adjustment execution amount (u') of each service level and the level cost (c), taking each product The minimum value of the sum is an optimal cost combination, and the optimal cost combination is not greater than the first execution cost. After that, the demand allocation amount (r) and the adjustment execution amount (u') of each service level can also be adjusted until the following conditions are met: [1] The sum of the several demand allocation amounts (r) is equal to the service demand. Quantity (q); [2] The adaptive execution amount (u') of each virtual machine is not greater than the single purchase amount (n); [3] Calculate the demand allocation amount (r) and the adjustment execution amount (u' for each service level. As the number of batches of each virtual machine, the product of the batch number of each service level and the average time (t) is calculated as the execution time of each service level, and the maximum value of each execution time is not greater than the minimum execution time. ;[4] calculate each service The sum of the adjustment execution amount (u') of the rank and the grade cost (c), the minimum value of each product sum is the above-mentioned optimal cost combination, and the optimal cost combination is not greater than the limit cost; [5] calculating each service level The demand allocation amount (r) and the modulus of the adjustment execution amount (u'), and the product of the modulus and the batch number of each service level is calculated as the first provisional value, and the batch number is singled at each service level. The purchase amount (n) is summed with the first provisional value, and multiplied by the adjustment execution amount (u') of each service level as the second provisional value, and the second provisional value and the average time (t) of each service level are calculated. The sum of the products, the minimum of the product sums is the best performance time described above.

Therefore, in an embodiment of the resource adaptation method of the cloud service of the present invention, the service level resource with high price can be adjusted to the service level resource with low price, thereby achieving the required service requirement, even if there is one or more on the same physical machine. Virtual machines can't work. They can also consider the virtual machine rental cost and overall performance. According to the change of the total demand of virtual machines, the number of virtual machines can be adjusted in different levels without affecting the operation of other virtual machines. "Cloud service quality" and "reduce cloud service cost" and other effects.

While the invention has been described in connection with the preferred embodiments described above, it is not intended to limit the scope of the invention. The technical scope of the invention is protected, and therefore the scope of the invention is defined by the scope of the appended claims.

S1‧‧‧ parameter setting procedure

S2‧‧‧First operation steps

S3‧‧‧Second operation steps

S4‧‧‧ third operation steps

Claims (5)

A resource adaptation method for a cloud service is implemented by a computer system, including: setting a limit cost, a service demand quantity, a level quantity, a plurality of service level level costs, an average time, a demand allocation amount, and a single execution quantity And a single purchase amount, the level cost, the average time, the demand allocation amount, the single execution amount, and the quantity of the single purchase quantity are all the quantity of the level; adjusting the demand allocation amount of each service level and single execution Quantity, until the following conditions are met: (1) The sum of the demand distributions is equal to the service demand, (2) the single execution quantity of each service level is not greater than the single purchase quantity, and (3) each calculation The quotient of the demand level of the service level and the quotient of the single execution amount is the number of batches of each service level, and the product of the batch quantity and the average time of each service level is calculated as the execution time of each service level, so that the largest of each execution time The value of the minimum is the minimum of each execution time, and the minimum is the minimum execution time. (4) The product of the single execution quantity of each service level and the level cost is not greater than the limit. Cost, setting the product sum of the single execution amount of each virtual machine and the level cost as a first execution cost; and changing the single execution amount of each virtual machine to an adjustment execution amount, adjusting the demand allocation amount of each service level and Adjust the execution amount until the following conditions are met: (1) The sum of the several demand allocations is equal to the service demand, (2) the adjustment execution amount of each virtual machine is not greater than the single purchase amount, and (3) the calculation The quotient of the demand distribution amount and the adjustment execution amount of each service level is used as the batch number of each virtual machine, and the product of the batch number and the average time of each service level is calculated as the execution time of each service level, and the largest of each execution time The value is not greater than the minimum execution time, (4) calculating the product sum of the adjustment execution amount of each service level and the level cost, and taking the minimum value of each product sum as an optimal cost combination, the optimal cost combination is not greater than the first Execution costs. According to the resource adaptation method of the cloud service described in claim 1 of the patent application scope, Including: adjusting the demand allocation of each service level and adjusting the execution amount until the following conditions are met: (1) the sum of the several demand allocations is equal to the service demand, and (2) the adjustment execution amount of each virtual machine (3) Calculate the demand distribution amount of each service level and the quotient of the adjustment execution amount as the batch quantity of each virtual machine, and calculate the product of the batch quantity and the average time of each service level as each service level. Execution time, the maximum value of each execution time is not greater than the minimum execution time, (4) calculating the product sum of the adjustment execution amount of each service level and the level cost, and the minimum value of each product sum is an optimal cost combination. The optimal cost combination is not greater than the limited cost, (5) calculating the demand allocation amount of each service level and the modulus of the adjustment execution amount, and calculating the product of the modulus and the batch number of each service level as a first provisional value. Calculate the number of single purchases of the batches at each service level and the first provisional value, and multiply the adjustment execution amount of each service level as a second provisional value to calculate the number of each service level. Provisional value and the product of the average time taken for each product and for the minimum time to an optimum performance. The resource adaptation method of the cloud service according to claim 1 or 2, wherein the calculation method of the minimum execution time is as follows: (1) and (2): Wherein, T _execute for the minimum execution time, l is the number of each class of service, L is the number of all the service levels, the total computation requirement for the service _{q, r 1, ..., r l} , ..., r L for The total amount of calculation of the service demand is the amount of allocation at each service level, and u ₁ , ..., u _l , ..., u _L are the amount of calculations actually performed each time the service level is performed. For the number of batches for each service level, For the execution time of each service level, n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level, and t ₁ , ..., t _l , ..., t _L are the service levels of each service level. The average time, c _l is the service cost of each level of virtual machine, C is the limit cost, and C' is the first execution cost. The resource adaptation method of the cloud service according to claim 1 or 2, wherein the calculation method of the optimal cost combination is as follows: Wherein, T _execute for the minimum execution time, l is the number of each class of service, L is the number of all the service levels, the total computation requirement for the service _{q, r 1, ..., r l} , ..., r L for The total amount of operation demand for service is allocated to each service level, and u ₁ ', ..., u _l ', ..., u _L ' is the amount of adjustment for each service level. For the number of batches for each service level, For the execution time of each service level, n ₁ , ..., n _l , ..., n _L are the single execution quantities currently purchased for each service level, and t ₁ , ..., t _l , ..., t _L are the service levels of each service level. The average time, c _l is the service cost of each level of virtual machine, C' is the first execution cost, and C' is the optimal cost combination. According to the resource adaptation method of the cloud service described in claim 2, wherein the optimal performance time is calculated as follows: Where T _{total is} the optimal performance time, _{Texem is} the minimum execution time, l is the number of each service level, L is the number of all service levels, q is the total operation amount of the service demand, r ₁ , ..., r _l , ..., r _{L are} the allocation amount of the total calculation amount of the service demand at each service level, and u ₁ ', ..., u _l ', ..., u _L ' are the adjustment execution amounts of the respective service levels, For the number of batches for each service level, For the execution time of each service level, n _l is the single execution quantity currently purchased by each service level, and t ₁ , ..., t _l , ..., t _L are the average time of each service level, For the first provisional value, For the second provisional value, C' is the first execution cost, and C' is the optimal cost combination.