CN110650032A - Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment - Google Patents
Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment Download PDFInfo
- Publication number
- CN110650032A CN110650032A CN201810682219.2A CN201810682219A CN110650032A CN 110650032 A CN110650032 A CN 110650032A CN 201810682219 A CN201810682219 A CN 201810682219A CN 110650032 A CN110650032 A CN 110650032A
- Authority
- CN
- China
- Prior art keywords
- service
- cloud
- deployment
- qos
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5041—Network service management, e.g. ensuring proper service fulfilment according to agreements characterised by the time relationship between creation and deployment of a service
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0893—Assignment of logical groups to network elements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5003—Managing SLA; Interaction between SLA and QoS
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5003—Managing SLA; Interaction between SLA and QoS
- H04L41/5019—Ensuring fulfilment of SLA
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5041—Network service management, e.g. ensuring proper service fulfilment according to agreements characterised by the time relationship between creation and deployment of a service
- H04L41/5054—Automatic deployment of services triggered by the service manager, e.g. service implementation by automatic configuration of network components
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/51—Discovery or management thereof, e.g. service location protocol [SLP] or web services
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of cloud computing and multi-cloud service, and relates to a method for constructing a QoS-based application optimization deployment scheme in a multi-cloud environment, which is developed aiming at an overall design framework, wherein the framework comprises the following steps: aiming at a user demand entity and a cloud service entity, corresponding entity models are designed to express configuration settings of application specifications of both parties, and SLA mechanism fuzziness and objective functions are solved. And modeling based on a QoS model aiming at the application optimization deployment problem in a multi-cloud environment, and providing constraint conditions and an objective function for optimization deployment. Meanwhile, different solutions are provided for the heuristic search algorithm aiming at the multi-objective optimization strategy applied by single cloud application deployment and cross-cloud application deployment requests. According to the invention, an effective application service optimization deployment scheme is designed, so that the automation level and the service quality level of the cloud service are improved.
Description
Technical Field
The invention belongs to the field of cloud computing and multi-cloud service, and particularly relates to a method for constructing a QoS-based application optimization deployment scheme in a multi-cloud environment.
Background
With the further development and wide application of cloud computing technology, various cloud service providers in the market are in the position. The cloud computing is prompted to enter a practical application stage as a new computing mode quickly due to the advantages of quick and accurate response, dynamic expandability, flexibility, convenience, reduction in operation cost and the like. On the one hand, as competition becomes more intense, quality of service (QoS) becomes more and more important, and although a service mechanism based on a Service Level Agreement (SLA) exists, specific parameters in the SLA and a provider do not give detailed explicit statements, so that a user does not know the actual meaning of the SLA, and the confidence of the user on the service is reduced. Meanwhile, the expression of the QoS by each large manufacturer in the SLA is not uniform, so that the user cannot compare and select among all the manufacturers. In the face of a complex market situation, how to purchase the most appropriate cloud service for the application demand of a user also becomes complex, and various influence factors such as different prices, different configurations and different service qualities influence the selection of the user. For cloud service providers, in order to maximize their economic benefits in a fierce market competition, they are gradually challenged more, which requires that service providers should improve service quality, thereby expanding their market share and realizing the increasing of economic benefits.
On the other hand, Multi-clouds (Multi-clouds) have been the subject of human enthusiasm in the last two years. In most cases, the use of multiple clouds, multi-type and multi-brand cloud deployments are not only logical, they can also provide better value than single cloud deployments. Within the industry, more and more enterprises are implementing development strategies using multiple cloud computing platforms to achieve, among other things, avoiding being limited to a certain vendor, increasing the ability to deliver available services, avoiding arbitrage differences, or maintaining control over certain sensitive information. In a cloudy scenario, a user may choose to use Amazon Web Services (AWS) simple storage service (S3) as storage, Rackspace one metal as a cloud database, Google as a big data system, and an OpenStack-based private cloud to manage sensitive data and applications. All of these resources work together to build one or more systems that allow enterprises to mix to meet a particular need using the services of a private cloud while meeting public needs.
In the face of a complex multi-cloud market, complex deployment requirements and various resource specifications, for a user, how to select an optimal deployment scheme meeting the requirements of the user is a troublesome and error-prone decision-making process. The current cloud service system has no corresponding service mechanism. If an agent mechanism is introduced between the user and the cloud service provider, the most economical and appropriate cloud service can be provided for the user, the complexity of the market is shielded for the user, the resource fragments of the cloud service provider can be reduced, the resource utilization rate is improved, and more SLAs are achieved. The introduction of the mechanism further perfects a cloud service system and improves the cloud service level. Based on the scheme, a Broker is added between a user and a cloud computing service provider, the Broker integrates various resource conditions of the current cloud computing market, communication negotiation is carried out between the Broker and the user, the user is helped to make a final optimized deployment decision, and the deployment decision problem is greatly solved for an inexperienced user who is not familiar with cloud computing.
Disclosure of Invention
The invention aims to provide a method for constructing an application optimization deployment scheme based on QoS (quality of service) in a multi-cloud environment, which is based on QoS, designs an integral framework and gradually expands the framework, and aims at a user demand entity and a cloud service entity to respectively describe the specific requirements of user application and services provided by a cloud service provider.
The framework of the invention comprises: aiming at a user demand entity and a cloud service entity, corresponding entity models are designed to express configuration settings of application specifications of both parties, and the current SLA mechanism fuzziness and objective functions are solved.
The technical scheme of the invention is specifically introduced as follows.
In a multi-cloud environment, a QoS-based application optimization deployment scheme mainly focuses on how to implement the multi-cloud environment, and automatically optimizes and deploys application services, and the implementation of the scheme needs to meet the SLA requirements of users based on QoS, and simultaneously guarantees that acceptable application performance is achieved, and the specific process is as follows:
(1) application deployment architecture design based on QoS requirements
In order to realize automatic service optimization combination in a multi-cloud environment, the invention realizes a set of universal and comprehensive-function intermediate agent architecture design. On one hand, the method receives the QoS request input of the user functionality and non-functionality, and enables a service combination matching algorithm on behalf of the user to provide an optimal solution for the user.
(2) A QoS solid model for the system is proposed
In order to solve the ambiguity of the current SLA-based pricing strategy, the system can conveniently carry out automatic negotiation and implement a service combination optimization matching algorithm, the request of a user and the delivery of a service are quantified, and a user request QoS entity and a service delivery entity are established. Wherein the measurement problem of the QoS parameters is mainly solved. The establishment of the QoS model effectively shields the disadvantages of the existing SLA mechanism, and the QoS data comes from the third-party detection institution CloudHarmony, which objectively defines a set of measurement standards from the perspective of users.
(3) An application deployment request model under a multi-cloud environment is provided
In a multi-cloud environment, in the face of various requests, a set of corresponding request models are needed to support the deployment request of the application, and meanwhile, a set of application deployment request models based on a graph in the multi-cloud environment is provided. Under a multi-cloud environment, a service combination matching algorithm is provided, and the most suitable cloud service combination for the user is selected for deployment.
Cloud computing expands the concept of resource sharing in a network-based business model, provides a customized computing environment for individuals and enterprise users through a simple access interface, is a service supply system, and greatly influences market development conditions on the quality of service based on a business model with service supply as a main body. In order to achieve the optimal effect, the invention integrates single cloud services provided by a plurality of clouds, constructs a cloud service system, selects the cloud service with the lowest cost and capable of meeting the QoS requirement of the cloud service, simultaneously improves the disaster tolerance capability of the system by multi-cloud deployment, and has great challenge on the management and operation of the cloud combined service in the multi-cloud environment, which is derived from the isomerism among different clouds and the complexity of the deployment relationship among the services.
Based on the processes, the invention designs a service optimization deployment scheme in a cloud environment, provides a service deployment scheme which achieves the maximum effect in the cloud environment, further promotes the development of cloud services, simultaneously carries out modeling description on the services from the perspective of users, shields complex market situation for the users, allows the users to pay attention to the service requirements concerned by the users, greatly improves the convenience of the users, and simultaneously carries out balance comparison between cloud service providers for the users objectively and fairly by using data of a third-party detection mechanism.
In the invention, the operation flow of the whole system is that firstly, a user puts forward specific service requirements according to the service attribute of the user, such as the number of server nodes, service software required to be installed on each node and the network connection condition among the nodes. Secondly, the service requirement which is most concerned by the user is that the user puts forward a corresponding service quality level, such as budget condition, service response capability, service availability and other specific service quality requirements. Secondly, the system models the user request through a service deployment request modeling module of the system according to the service deployment situation proposed by the user, and finally forms a deployment request graph G (V, E), wherein the nodes represent service nodes required by the user, and edges in the graph are the network connection situation between the nodes, for example, data transmission between an application service logic node and a database node. Next, the system calls a QoS modeling module and a cost budgeting module to perform modeling according to the specific QoS requirement and the budget condition proposed by the user, and provides decision basis data for the next optimization deployment decision.
In the invention, the system calls the detailed information of the cloud service provider entity stored in the persistence library, carries out filtering and screening according to the QoS and the budget model, and then calls a heuristic backtracking algorithm to find an optimized service deployment scheme.
In the invention, each cloud provider can provide cloud services with different types and different service qualities. Then service deployment in a multi-cloud environment would take into account these cloud service provider mockups.
In view of the above problems, the solution requires the following steps in total:
first, we give relevant definitions and metrics of relevant properties, availability, reliability, and responsiveness for the QoS model. Availability refers to the percentage of uptime, t and t, of a cloud service over a period of timesRespectively representing the normal operation time and one operation period, the calculation formula is represented as:
reliability refers to the guarantee that the cloud service is free from hardware faults, software error reporting and other conditions which can cause cloud service downtime, and n aresRespectively representing the operation causing the cloud service to be down and the total number of operations in a time period, the reliability can be represented as
The response capability refers to the corresponding speed of the cloud service to the request within a certain time interval. t is tiRepresenting the time interval from submission to completion of the ith request, n being the number of requests submitted in a certain time period, tmaxIndicating an acceptable maximum response time (t)i≤tmax) Then the response capability can be expressed as:
secondly, aiming at the multi-cloud service provider model, a, vm and Cdata are usedin(Pk)，Cdataout(Pk) Respectively represent service providers PkProvided are a virtual application device, a virtual machine, a cost of data transmission between internal networks, and a cost of data transmission between external networks. Virtual device application a mayTo represent with four elements: the specific mathematical formula expressed by the application type, price, license certificate type and application device size is as follows: a { ApplianceType; cost; license type; size }. The virtual machine vm, can be represented by two elements: virtual machine type and price. The specific formula is expressed as: vm: { MachineType; cost }.
Finally, for the user service deployment request model, it can be abstracted into a graph structure G (V, E), and the vertices in the graph represent the server nodes (virtual machines running application devices) required by the user application deployment, and can be expressed by the following formula: sv ═ appliance, virtualunit ═ av,vmv},
The edge e { v, v '} indicates that the server nodes v and v' are communicated with each other, and the data traffic between the connected vertices is indicated by "D". For service node Sv, assume that its appliance is leased from PkVirtual machine renting from PlThen, for the cost of Sv, the cost of leasing appliance (av, pk) and the cost of leasing virtual machine (vmv, pl) are included, if the two are not from the same service provider and the data transmission cost of appliance is added, the calculation formula of the total cost can be expressed as:
when service node Sv ═ { a ═ av,Pk,vmv,plJ and Sv' ═ av',Pk',vmv',pl'When there is a network connection between nodes, the cost of renting occupied network resources between nodes can be expressed as:
so all deployment costs for one user can be expressed as:
in the invention, in a cloud computing environment, because a plurality of influence factors are considered and the cloud computing environment faces wider and wider market conditions, the problem is a complex decision problem with a larger problem space, and for the cloud computing environment, a related decision algorithm is designed, and finally a cloud service selection matching algorithm based on effectiveness is provided, so that the provided service optimization deployment problem is solved, wherein the service optimization deployment problem comprises a heuristic backtracking algorithm and a particle swarm optimization algorithm based on heuristic search. In a multi-cloud environment, users may have different QoS requirements for a certain service deployment scheme (or for a certain cloud service provider), and for each QoS requirement, it is desirable that the larger the value is, the better the QoS value is (the better the service attribute is), or the smaller the value is, the better the service attribute is (the larger the QoS value is, the better the service attribute is). The system is a multi-objective optimization problem, and an optimal service deployment scheme needs to be selected for a user from a plurality of schemes according to the actual requirements of the user. For a multi-objective optimization problem with L objective functions and several constraints, the mathematical model can be expressed as a vector of n-dimensional Euclidean space, the multi-objective optimization is the optimization of vector functions, and relatively speaking, a single objective function is the optimization of scalar functions.
The degree of satisfaction of the user with the service provided by the cloud service provider is represented by the calculation of the efficacy function. Combining a multi-objective optimization problem model, for each QoS service quality, a corresponding efficacy function for measuring the satisfaction degree of a user should be provided, in the multi-objective optimization problem model, an index function is adopted to simulate the preference degree of the user to a certain service attribute, a common efficacy function is provided according to the actual situation of the user, the lower the user aims at the service attribute, the better the user aims at the service attribute, the efficacy function is a decreasing gargle, and the higher the better the service attribute, the efficacy function is an increasing function. For service attributes with lower values and better, the utility value can be expressed as:
wherein u is1Indicating utilityThe value, x, denotes the service attribute (0. ltoreq. x. ltoreq.1) as small as possible1，b1Is a constant positive value, k1Is such that u1(0)＝1，u1(1) A scale factor of 0. Since the final evaluation function is a linear weighting function, our goal is to achieve the scheme that maximizes the overall work function, so the choice of specific parameters (a) satisfies the above constraints1,b1Choice of) has no effect on our end result, since we use the relative magnitude of the value under different scenarios, and do not intend what its specific value is. Thus, let a1Can be derived from 1
Likewise, for a service attribute that is better for a higher value, its utility function can be expressed as:at this time let a2Can be derived from 1
And then, building an efficacy model for the multi-attribute values by using a weighted summation function, wherein the comprehensive utility value can be expressed as:
wherein u is3Representing the overall utility value, p representing a proposal, wjRepresenting the attribute weight, uiIs an attribute, xjIs a utility function since u1(x)，u2(x) Is a monotonic function, we can easily map attribute values to their corresponding utility functions. At the same time, since u3(p) is stackable and we can easily extend it to any number of service attributes we need. We also propose a normalized QoS model based on actual comparison of QoS quantities to give a more objective service level, and thus select the best service provider. At the same time, the normalized QoS value is used to express the aggregate QoS value of single component service, and then each single group is further integratedThe aggregate QoS value of a service represents an objective function of the combined service QoS utility. Cross-cloud service combined objective function u still adopts weighted decision model u3Wherein u isi(xj) Replacing the normalized QoS attribute value, and simultaneously performing normalization processing to make the attribute value operate in the same range, wherein the normalization processing process is described by the following formula:
in both formula one and formula two, xiDenotes the ith QoS attribute value, ximaxAnd ximinRespectively representing the maximum and minimum values of the QoS attribute value in the candidate service provider, if the QoS attribute value is a forward attribute, adopting a formula I to perform normalization calculation, otherwise adopting a formula II to perform calculation, and finally substituting the normalized value into the formula IIThe calculation is carried out, and if the calculated value is larger, the strategy is more suitable for the user.
In the invention, a heuristic optimization search method, namely a particle swarm optimization algorithm, can be effectively solved, and the application of single cloud deployment and cross-cloud application deployment specifically comprise the following steps:
due to special application requirements or preferences of users, the users may select to deploy applications in a single cloud environment, which faces the selection problem of multiple application service providers, aiming at single cloud deployment, an enumeration algorithm can be selected to perform optimal deployment strategy selection, and the optimal solution can be solved within O (N) (the number of service providers) time complexity, and the enumeration algorithm can obtain the optimal solution. The cross-cloud application deployment is carried out by adopting a particle swarm optimization algorithm according to a user application service deployment request model and a cross-cloud deployment problem which is a multi-target optimal path problem under a multi-constraint condition because service combination calculation is requiredThe row solution is deployed across cloud applications, n application service deployment nodes (server nodes) are arranged, and n are arranged in the ith service groupiA candidate service provider, SGi＝(si1，…，sini)，sijThe number of a particular service provider in the ith service group. The cross-cloud composite service scheme forms an n-dimensional vector S ═ (S)1i，…，sni) Wherein s isij∈[1，ni]In the basic particle swarm optimization, the speed and displacement model aims at continuous numerical solution, while the combined optimization problem involved in our invention is a discrete problem,
v [ i +1] (int) w v [ i ] + c1 r1 (pbest [ i ] -present [ i ]) + c2 r2 (gbest-present [ i ]) (formula three)
present [ i +1] ═ present [ i ] + v [ i ] (formula four)
In equations three and four, r1 is a random number such that c1 r1 (pbest [ i ] -present [ i ]) can take each integer within the range of [0, c1 (pbest [ i ] -present [ i ]) ] (or [ c1 (pbest [ i ] -present [ i ]),0]), etc., and r2 is similar. The process of the algorithm is described as follows:
step1 initializing a particle group of size M, setting initial position and depth
step2 calculating fitness value of each particle
step3 comparing the fitness value of the particle with the fitness value of its best position pbest [ i ], and taking the best value
step4 comparing the fitness value of each particle with the fitness value of the global best position gbest, and taking the best as the global optimum
step5, according to the velocity formula and the position formula, the range of the new generation position Xi and the velocity Vi is checked, the search range and the maximum velocity of the particle are limited, and the velocity and the position of the particle are updated.
step6 output if the termination condition is satisfied, otherwise return to step2 (condition: output result can satisfy predetermined minimum adaptation threshold)
In addition, by designing a service selection method, candidate services with poor QoS are eliminated, so that the scale of the candidate services is reduced, the service selection efficiency is improved, and for each service node, the service selection can be considered to be performed only in the filtered service set of the corresponding service group.
Compared with the prior art, the invention has the beneficial effects that:
the cloud service system and the cloud service method have the advantages that in a multi-cloud environment, the most economical and appropriate cloud service can be provided for users, the complexity of the market is shielded for the users, the resource fragments of cloud service providers can be reduced, the resource utilization rate is improved, more SLAs are achieved, the cloud service system is further improved, the cloud service level is improved, and the users are helped to make final optimized deployment decisions through the application optimized deployment scheme in the multi-cloud environment.
Drawings
Fig. 1 is a flow chart of a prototype system.
Fig. 2 is a diagram of an application service topology.
FIG. 3 is a graph showing the results of the experiment.
FIG. 4 is a graph showing the results of the second experiment.
FIG. 5 is a graph showing the results of the three experiments.
FIG. 6 is a diagram of pre-copy versus direct copy.
FIG. 7 is a diagram of the overall architecture design for multi-cloud service deployment in accordance with the present invention
Detailed Description
The technical solution of the present invention is specifically described below with reference to the accompanying drawings and examples.
The invention aims to provide a QoS-based application optimization deployment scheme under a multi-cloud environment. As shown in fig. 7, the application optimization deployment system based on the multi-cloud architecture includes a user interface, a cloud service agent layer, and a cloud service provider. In the architecture design, all application services provided for the user are presented to the user through a web interface, and this part provides a convenient imaging interactive interface for the user, so that the user can conveniently put forward user requirements, such as software deployment, hardware, QoS requirements (such as maximum acceptable network delay, cost budget, reliability, and the like), and meanwhile, mechanisms such as user management and the like are realized, and user authentication login and user operation record recording and the like are realized. The cloud service agent layer is the key point of the design of the invention, and mainly comprises a multi-cloud service deployment request model, a QoS measurement model, a service combination optimization decision in a multi-cloud environment, service deployment management, SLA monitoring management, a cloud service API and QoS negotiation extension. And the cloud service provider undertakes corresponding service deployment and service output, such as public cloud services provided by Amazon, Microsoft and Google.
Example 1
In the invention, the QoS-based application optimization deployment scheme under the cloud environment is specifically divided into the following three stages:
1. application deployment system design based on QoS requirements
The content of the application deployment architecture design based on the QoS requirement is elaborated here, and mainly starts from a service deployment request, firstly, a user proposes specific service requirements according to the service attributes of the user, such as the number of server nodes, secondly, a system mutes the user request according to the service deployment condition proposed by the user, and finally, a deployment request graph is formed, wherein the nodes represent all service nodes required by the user, the edges in the graph are the network connection condition between the nodes, and then, the system calls a modeling module and a cost budget module to mute according to the specific requirements and the budget condition proposed by the user, so as to provide decision-based data for the next optimized deployment. And then the system enters a core module and a service optimization deployment decision module, in the module, the system can use the detailed information of the cloud service provider entity stored in the persistence library, filter and select according to the QoS and budget model, and then call a heuristic backtracking algorithm to find an optimized service deployment scheme. And returning the proposed scheme to the user, obtaining a deployment opinion of the user, and then directly deploying according to a final decision of the user, or recalculating an optimized deployment scheme according to further adjustment of the user. The specific system design flow is shown in the first drawing.
2. QoS mockup design for a system
The user has different QoS requirements for a certain service deployment scheme, and for each QoS requirement, it is a target optimization problem that the larger the expected value is, the better the QoS value is (the better the service attribute is), or the smaller the expected value is, the better the QoS value is (the better the service attribute is the larger the QoS value is). For the QoS model, analysis, availability, reliability, responsiveness are mainly performed from several properties. The specific analysis is as follows:
(1) availability
The availability refers to the percentage of the normal operation time in a period of time of the cloud service, and the calculation formula is as follows:
where 0. ltoreq. a.ltoreq.1, denotes availability, t and tsRespectively representing the uptime and one run cycle. The closer a is to 1, the higher the availability is. Since the cloud service is provided through the internet, the network may be interrupted, and the user desires a service having high availability. Ideally, cloud services should be run uninterrupted.
(2) Reliability of
The reliability is a guarantee that the cloud service is free from hardware faults, software error reporting and other faults which can cause the cloud service to be down, and a specific calculation formula is as follows:
where r is 0. ltoreq. r.ltoreq.1, denotes reliability, n and nsRespectively, the operation causing the cloud service to be down, and the total number of operations in one time period. The closer the value of r is to 1, the higher the surface reliability. Since cloud services rely on hardware facilities and related software, it is not clear which part will fail and users desire greater reliability. Ideally, cloud services should be fault-free.
(3) Responsiveness capability
The response capability refers to the corresponding speed of the cloud service to the request within a certain time interval. Can be calculated by the following formula:
where 0. ltoreq. tau. ltoreq.1 denotes the response capability, tiRepresenting the time interval from submission to completion of the ith request, n being the number of requests submitted in a certain time period, tmaxIndicating an acceptable maximum response time (t)i≤tmax) F is a function of statistical overall response trend, and can be an average function, a median function, and the like. The closer the value of τ is to 1, the stronger the service response capability is.
3. Application deployment request model in multi-cloud environment
In summary, cloud service deployment comprises two stages, namely service discovery, namely collecting different cloud service providers in a cloud computing market, recording service types and specifications provided by the different cloud service providers, and providing selection objects for a later application deployment scheme. The second stage is a service optimization selection stage, and the second stage integrates the information collected in the first stage, and selects the most suitable service resource and provider thereof for the user application under the guidance of an optimization algorithm. The wide variety of cloud service resources and user-specific application quality of service requirements make the process of selecting the optimal service combination more complex and difficult.
The first step of the application optimization deployment scheme in the cloud environment is to establish a model for cloud resources and user QoS requirements, and the model can be solved based on the existing cloud service components and the QoS model provided by the invention. The next step is how to select the most suitable cloud service (single cloud deployment) or combined cloud service (cross-cloud deployment) to ensure that the service deployment cost of the user is the lowest and the service deployment efficiency is the highest while meeting the functional and non-functional requirements of the user aiming at different user requests.
Example 2
One-cloud environment and multi-cloud environment optimal cloud service provider selection
In the experiment, a typical three-layer WEB application service deployment structure is used as a request example of application deployment, and a WEB system is supposed to be deployed by a current user, wherein the system comprises two WEB front-end servers, three business logic processing servers and two database servers. In order to simplify the model, it is assumed in the experiment that the applications used by the application servers are all open-source and free, and each cloud service manufacturer has a corresponding virtual image, so in this case, the lease cost of the service node only includes the lease cost for the VM and the data transmission cost between the nodes.
According to the service deployment request, the 7 service machines are respectively numbered as 0-6, and the network connection situation and the data transmission situation are shown in fig. two: (the number is the average transmission rate GB/H, null indicates no data transmission)
According to the specific requirements of the user, the user can give specific value ranges and weights aiming at various QoS parameters, for example, in the experiment, the reliability weight is defined to be 0.4, the response capacity is 0.2, the budget weight is 0.4 and the like, the requirements and the emphasis points of the user on the specific service quality are shown, corresponding QoS parameter values are calculated aiming at a QoS model provided by each service provider, then the QoS values are used for screening the cloud service providers meeting the requirements of the user, and meanwhile, a candidate queue of the cloud service providers is provided and used as the input of a decision algorithm.
In the experiment, three sets of Openstack environments are deployed, different pricing and configuration information is given out according to the current market condition, monitoring data given out by simulating CloudHarmony is captured, and each cloud service provider is calculated according to the model. In order to simulate the diversity of the real market, in the experimental environment, Openstack1 is relatively more economical, Openstack2 is relatively more excellent in performance, and Openstack3 is relatively more stable through setting specific monitoring data.
In order to make the experimental result more obvious, only part of QoS parameters are selected, in the invention, three experiments are respectively carried out, three times of requirement data of the three experiments are respectively input into a Broker system, the Broker system immediately returns an optimal service provider which is Openstack3, and relevant parameters of Openstack3 and comparison with user requirement parameters are given, in the three experiments, efficacy comparison of three deployment schemes is respectively shown in a third graph, a fourth graph and a fifth graph. Through experiments, the overall efficacy value is not very different, but the Broker system can provide cloud service providers which meet the requirements of users to a certain extent through the wishes of the users.
Second, cross-cloud service deployment experiment verification
In the experiment of the invention, in the particle swarm optimization algorithm, w in the velocity updating formula is called as the inertia weight, so that the inertia of the particles is kept in the flying process, and the particles can keep continuously exploring a part of space. If w is 0, the particle velocity has no memory and always flies to the weighted center of the individual optimal position and the global optimal position, and the algorithm easily falls into the local optimal solution under the condition. when w is increased, the particles have the capability of global search, and accordingly the balance between the global search capability and the local search capability can be adjusted by adjusting w according to specific problems. In general, it works well to have w decrease stepwise from 1.4 to 0 during each iteration. c1 and c2 are individual and global learning factors, also called acceleration factors, respectively, representing the individually and globally optimal acceleration weights of the particle in flight, and are typically set to 2 empirically. When c1 is 0, the particles do not have individual cognitive ability, and the learning of the population is emphasized, and the convergence can be faster, but the particles easily fall into a locally optimal solution. When c2 is 0, particles are searched only by individual experience, and the optimal solution cannot be obtained in some cases without learning population.
In the invention, the particle swarm optimization algorithm is verified by finding the process of the function y being 1-cos (3 x) exp (-x) at the maximum value of [0,4 ]. Firstly, a standard particle swarm function (continuous) is utilized to calculate the value of the function, secondly, the value of the function is calculated by limiting the displacement unit to be an integer (discrete), and the result of the function is verified to be the correctness and the effectiveness of the particle swarm algorithm provided by the invention through comparison, and as shown in the sixth figure, the result of the function can be shown, based on the QoS requirement of a service user, two Web-front servers (server0-server1) are deployed on an Openstack1 simulation cloud service provider, and three application servers (server2-server4) and two database servers are deployed in an Openstack3 simulation environment. The deployment scheme has the advantages of lowest cost and relatively high availability under the condition of meeting the QoS requirement of a user. The availability obtained by the user is higher while the budget is lower than that obtained by the user when deployed solely on the Openstack3 environment.
Claims (6)
1. A method for constructing a QoS-based application optimization deployment scenario in a cloud environment, comprising,
(1) application deployment system design based on QoS requirements
Providing a universal and comprehensive-function intermediate agent architecture design, receiving the functional and non-functional QoS request input of a user, representing the benefit of the user, starting a service combination matching algorithm, and providing an optimal solution for the user;
(2) proposing a QoS solid model for the system
Quantifying the request of a user and the delivery of service, and establishing a user request QoS entity and a service delivery entity; solving the measurement problem of the QoS parameters; the method can be used for shielding the defects of the existing SLA mechanism and objectively defining a set of measurement standards from the perspective of a user;
(3) application deployment request model under multi-cloud environment
Establishing a set of corresponding request models to support the deployment request of the application, simultaneously establishing a set of application deployment request models based on a graph in a multi-cloud environment, providing a service combination matching algorithm in the multi-cloud environment, and selecting the most appropriate cloud service combination for deployment.
2. The method according to claim 1, wherein the operation flow of the constructed scheme includes that, firstly, a user proposes specific service requirements according to own service attributes, including: the number of server nodes, the service software required to be installed on each node and the network connection condition among the nodes are calculated; secondly, the user puts forward the corresponding service quality level, such as budget condition, service response capability and service quality requirement of service availability, then the scheme models the user's request through the service deployment request modeling module of the scheme system according to the service deployment condition put forward by the user, finally forms a deployment request graph G (V, E), wherein the nodes represent each service node required by the user, and finally. The scheme system calls a QoS modeling module and a cost budget module to model according to the specific QoS requirement and budget condition provided by the user, and provides decision basis data for the optimized deployment decision.
3. The method of claim 2, wherein the scheme system invokes details of cloud service provider entities stored in the persistence library, filters based on QoS and budget models, and then invokes a heuristic backtracking algorithm to find an optimized service deployment scheme.
4. The method according to claim 2 or 3, wherein different kinds of cloud services with different service qualities are provided for the cloud provider, and a cloud service provider entity model is provided by the following steps:
firstly, relevant definition and measurement standards of relevant properties, availability, reliability and responsiveness are given for a QoS model, wherein the availability refers to the percentage of the normal operation time, t and t in a period of time of a cloud servicesRespectively representing the normal operation time and one operation period, the calculation formula is represented as:
reliability refers to the guarantee that the cloud service is free from hardware faults, software error reporting and other conditions which can cause cloud service downtime, and n aresRespectively representing the operation causing the cloud service to be down and the total number of operations in a time period, the reliability can be represented as
The response capability refers to the corresponding speed, t, of the cloud service to the request within a certain time intervaliRepresenting the time interval from submission to completion of the ith request, n being the number of requests submitted in a certain time period, tmaxIndicates the most acceptableLong response time (t)i≤tmax) Then the response capability can be expressed as:
secondly, aiming at the multi-cloud service provider model, a, vm and Cdata are usedin(Pk)，Cdataout(Pk) Respectively represent service providers PkThe virtual application device, the virtual machine, the cost of data transmission between internal networks and the cost of data transmission between external networks are provided, and the virtual device application a is represented by four elements: the specific mathematical formula expressed by the application type, price, license certificate type and application device size is as follows: a { ApplianceType; cost; license type; size };
virtual machine vm, represented by two elements: the type and price of the virtual machine are expressed by the following specific formula: vm: { MachineType; cost };
finally, abstracting the user service deployment request model into a graph structure G (V, E), wherein the vertex in the graph represents a server node (a virtual machine for running an application device) required by user application deployment, and the vertex is expressed by a formula:
the edge e { v, v '} indicates that the server nodes v and v' are communicated with each other, and the data traffic between the connected vertices is indicated by "D". For service node Sv, assume that its appliance is leased from PkVirtual machine renting from PlThen, for the cost of Sv, including the cost of leasing appliance (av, pk), and the cost of leasing virtual machines (vmv, pl), if the two are not from the same service provider, plus the data transmission cost of appliance, the calculation formula of the overall cost is expressed as:
when service node Sv ═ { a ═ av,Pk,vmv,plJ and Sv' ═ av',Pk',vmv',pl'When network connection exists between nodes, the cost of renting the occupied network resources between the nodes is represented as:
DSize(e)*CDatain(pl)*T if l＝l'
so all deployment costs for one user are expressed as:
5. the method according to claim 1 or 5, wherein for the cloud computing environment, a related decision algorithm is designed, and finally a utility-based cloud service selection matching algorithm is given, wherein the algorithm comprises a heuristic backtracking algorithm and a particle swarm optimization algorithm based on heuristic search; wherein the content of the first and second substances,
the satisfaction degree of the user on the service provided by the cloud service provider is represented by calculation of a power function, wherein an exponential function is adopted to simulate the preference degree of the user on a certain service attribute, such as a common power function, the lower the user aims at the service attribute, the better the user aims at the service attribute, the power function is a decreasing gargle, the higher the better the service attribute, the power function is an increasing function, and the lower the value of the service attribute, the better the service attribute, the utility value is represented as:
wherein u is1Represents a utility value, x represents a service attribute (0. ltoreq. x. ltoreq.1) whose value is as small as possible, a1，b1Is a constant positive value, k1Is such that u1(0)＝1，u1(1) A scale factor of 0; the final evaluation function is a linear weighting function, let a1Can be derived from 1
For a service attribute with a higher value, the better, its utility function is expressed as:
at this time let a2Can be derived from 1
And then establishing an efficacy model for the multi-attribute values by using a weighted summation function, wherein the comprehensive utility value can be expressed as:
wherein u is3Representing the overall utility value, p representing a proposal, wjRepresenting the attribute weight, uiIs an attribute, xjIs a utility function since u1(x)，u2(x) Is a monotonic function that can map attribute values to their corresponding utility functions; at the same time, since u3(p) are superposable, and adopt and represent the aggregate QoS value of the single assembly service on the basis of the normalized QoS model, and use the normalized QoS value, and then further synthesize the aggregate QoS value of each single assembly service and represent the objective function of the QoS utility of the combined service; cross-cloud service combined objective function adopting weighted decision model u3Wherein u isi(xj) Replacing the normalized QoS attribute value, and simultaneously performing normalization processing to make the attribute value operate in the same range, wherein the normalization processing process is described by the following formula:
in both formula one and formula two, xiDenotes the ith QoS attribute value, ximaxAnd ximinRespectively representing the QoS attribute values of candidate service providersMaximum and minimum, if the QoS attribute value is forward attribute, adopting formula one to make normalization calculation, otherwise adopting formula two to make calculation, finally substituting the above-mentioned normalized value into
The calculation is carried out, and if the calculated value is larger, the strategy is more suitable for the user.
6. The method of claim 5, wherein a heuristic optimization search method-particle swarm optimization algorithm is employed, and wherein applying single cloud deployment and cross-cloud application deployment comprises:
the user can select to deploy the application in a single cloud environment, an enumeration algorithm is selected for optimal deployment strategy selection aiming at the single cloud deployment, the solution is realized within O (N) (the number of service providers) time complexity, and the enumeration algorithm can obtain an optimal solution;
deployment across cloud applications: the particle swarm optimization algorithm is adopted to solve the cross-cloud application deployment, n application service deployment nodes (server nodes) are arranged, and n are arranged in the ith service groupiA candidate service provider, SGi＝(si1，…，sini)，sijThe number of a particular service provider in the ith service group. The cross-cloud composite service scheme forms an n-dimensional vector S ═ (S)1i，…，sni) Wherein s isij∈[1，ni]In the basic particle swarm optimization, the velocity, displacement model is aimed at a continuous numerical solution, where the involved combinatorial optimization problem is a discrete problem,
v [ i +1] (int) w v [ i ] + c1 r1 (pbest [ i ] -present [ i ]) + c2 r2 (gbest-present [ i ]) (formula three)
present [ i +1] ═ present [ i ] + v [ i ] (formula four)
In equations three and four, r1 is a random number such that c1 r1 (pbest [ i ] -present [ i ]) can take each integer within the range of [0, c1 (pbest [ i ] -present [ i ]) ] (or [ c1 (pbest [ i ] -present [ i ]),0]), etc., and r2 is similar. The process of the algorithm is described as follows:
step1 initializing a particle group of size M, setting initial position and depth
step2 calculating fitness value of each particle
step3 comparing the fitness value of the particle with the fitness value of its best position pbest [ i ], and taking the best value
step4 comparing the fitness value of each particle with the fitness value of the global best position gbest, and taking the best as the global optimum
step5, according to the velocity formula and the position formula, carrying out range check on the newly generated position Xi and the velocity Vi, limiting the search range and the maximum velocity of the particles, and updating the velocity and the position of the particles;
step6 output if the termination condition is satisfied, otherwise return to step2 (condition: output result can satisfy predetermined minimum adaptation threshold)
Further comprising: by designing a service selection method, candidate services with poor QoS are eliminated, the scale of the candidate services is reduced, the service selection efficiency is improved, and for each service node, service selection can be considered to be performed only in a filtered service set of a corresponding service group.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810682219.2A CN110650032A (en) | 2018-06-27 | 2018-06-27 | Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810682219.2A CN110650032A (en) | 2018-06-27 | 2018-06-27 | Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110650032A true CN110650032A (en) | 2020-01-03 |
Family
ID=68988864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810682219.2A Pending CN110650032A (en) | 2018-06-27 | 2018-06-27 | Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110650032A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111835516A (en) * | 2020-06-14 | 2020-10-27 | 西安电子科技大学 | Public key repudiatable encryption method and system |
| CN112765407A (en) * | 2020-12-30 | 2021-05-07 | 重庆邮电大学 | QoS service combination method based on user preference in Internet of things environment |
| CN112769942A (en) * | 2021-01-15 | 2021-05-07 | 重庆邮电大学 | QoS-based micro-service dynamic arranging method |
| CN112926262A (en) * | 2021-02-18 | 2021-06-08 | 同济大学 | Data separate storage method, system, medium and terminal under cloud edge collaborative environment |
| CN114037340A (en) * | 2021-11-24 | 2022-02-11 | 天津大学 | Service attraction model-based cross-boundary service efficiency evaluation method |
| CN114095500A (en) * | 2021-11-09 | 2022-02-25 | 龙思云(北京)科技有限公司 | Non-inductive cross-region drift method and system for cloud application |
| CN114466057A (en) * | 2020-10-21 | 2022-05-10 | 神州数码信息系统有限公司 | Automatic optimization method for service configuration based on space search |
| CN117056043A (en) * | 2023-07-27 | 2023-11-14 | 广东工业大学 | Service selection arrangement method and system for SCOS |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103473320A (en) * | 2013-09-12 | 2013-12-25 | 南京大学 | Method for service combination facing cloud-spanning platform |
| CN103490938A (en) * | 2013-10-15 | 2014-01-01 | 河海大学 | Layering-based cloud service combination failure recovery system and method |
| CN104360908A (en) * | 2014-10-31 | 2015-02-18 | 东北大学 | Ant colony optimization algorithm-based SBS (service-based software system) resource allocation method in cloud environment |
| CN106357796A (en) * | 2016-10-12 | 2017-01-25 | 四川用联信息技术有限公司 | Optimal service allocation algorithm for mobile applications under mobile cloud computing |
| CN106357739A (en) * | 2016-08-27 | 2017-01-25 | 浙江工业大学 | Two-stage composition and scheduling method specific to lot-sizing cloud service request |
| CN106779283A (en) * | 2016-11-14 | 2017-05-31 | 阿坝师范学院 | A kind of SLA ensures the desired cloud computing service income methods of QoS |
-
2018
- 2018-06-27 CN CN201810682219.2A patent/CN110650032A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103473320A (en) * | 2013-09-12 | 2013-12-25 | 南京大学 | Method for service combination facing cloud-spanning platform |
| CN103490938A (en) * | 2013-10-15 | 2014-01-01 | 河海大学 | Layering-based cloud service combination failure recovery system and method |
| CN104360908A (en) * | 2014-10-31 | 2015-02-18 | 东北大学 | Ant colony optimization algorithm-based SBS (service-based software system) resource allocation method in cloud environment |
| CN106357739A (en) * | 2016-08-27 | 2017-01-25 | 浙江工业大学 | Two-stage composition and scheduling method specific to lot-sizing cloud service request |
| CN106357796A (en) * | 2016-10-12 | 2017-01-25 | 四川用联信息技术有限公司 | Optimal service allocation algorithm for mobile applications under mobile cloud computing |
| CN106779283A (en) * | 2016-11-14 | 2017-05-31 | 阿坝师范学院 | A kind of SLA ensures the desired cloud computing service income methods of QoS |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111835516A (en) * | 2020-06-14 | 2020-10-27 | 西安电子科技大学 | Public key repudiatable encryption method and system |
| CN111835516B (en) * | 2020-06-14 | 2021-11-23 | 西安电子科技大学 | Public key repudiatable encryption method and system |
| CN114466057A (en) * | 2020-10-21 | 2022-05-10 | 神州数码信息系统有限公司 | Automatic optimization method for service configuration based on space search |
| CN112765407A (en) * | 2020-12-30 | 2021-05-07 | 重庆邮电大学 | QoS service combination method based on user preference in Internet of things environment |
| CN112765407B (en) * | 2020-12-30 | 2022-11-11 | 重庆邮电大学 | QoS service combination method based on user preference in Internet of things environment |
| CN112769942A (en) * | 2021-01-15 | 2021-05-07 | 重庆邮电大学 | QoS-based micro-service dynamic arranging method |
| CN112769942B (en) * | 2021-01-15 | 2022-09-16 | 重庆邮电大学 | QoS-based micro-service dynamic arranging method |
| CN112926262A (en) * | 2021-02-18 | 2021-06-08 | 同济大学 | Data separate storage method, system, medium and terminal under cloud edge collaborative environment |
| CN114095500A (en) * | 2021-11-09 | 2022-02-25 | 龙思云(北京)科技有限公司 | Non-inductive cross-region drift method and system for cloud application |
| CN114037340A (en) * | 2021-11-24 | 2022-02-11 | 天津大学 | Service attraction model-based cross-boundary service efficiency evaluation method |
| CN114037340B (en) * | 2021-11-24 | 2024-06-04 | 天津大学 | Cross-border service effectiveness evaluation method based on service attraction model |
| CN117056043A (en) * | 2023-07-27 | 2023-11-14 | 广东工业大学 | Service selection arrangement method and system for SCOS |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110650032A (en) | Method for constructing QoS-based application optimization deployment scheme in multi-cloud environment | |
| Jalaparti et al. | Dynamic pricing and traffic engineering for timely inter-datacenter transfers | |
| Kochovski et al. | Smart contracts for service-level agreements in edge-to-cloud computing | |
| Chen et al. | The handbook of engineering self-aware and self-expressive systems | |
| Aryal et al. | Dynamic application deployment in federations of clouds and edge resources using a multiobjective optimization AI algorithm | |
| US20080162232A1 (en) | Method and apparatus for business process analysis and optimization | |
| De Maio et al. | Multi-objective mobile edge provisioning in small cell clouds | |
| CN109951333A (en) | Trust evaluation device based on subjective logic in the processing of edge calculations network video | |
| CN104038540A (en) | Method and system for automatically selecting application proxy server | |
| Anastasi et al. | Qbrokage: A genetic approach for qos cloud brokering | |
| Xu et al. | Cost-aware resource management for federated clouds using resource sharing contracts | |
| CN112199189A (en) | Adaptation of deep learning models to resource constrained edge devices | |
| Zhang et al. | IoV scenario: Implementation of a bandwidth aware algorithm in wireless network communication mode | |
| Avasalcai et al. | Resource management for latency-sensitive IoT applications with satisfiability | |
| Sharghivand et al. | A comprehensive survey on auction mechanism design for cloud/edge resource management and pricing | |
| CN113346938A (en) | Edge computing resource fusion management method for air-space-ground integrated network | |
| Xu et al. | Optimized contract-based model for resource allocation in federated geo-distributed clouds | |
| Hasenburg et al. | Supporting the evaluation of fog-based IoT applications during the design phase | |
| Li et al. | TSLAM: a trust-enabled self-learning agent model for service matching in the cloud market | |
| EP3346382A1 (en) | Forming neighborhood groups from disperse cloud providers | |
| Toumi et al. | On using deep reinforcement learning for multi-domain SFC placement | |
| Mezni et al. | A negotiation‐based service selection approach using swarm intelligence and kernel density estimation | |
| Sultana et al. | Two-tier architecture for spectrum auction in SDN-enabled cloud radio access network | |
| Kumar et al. | A user-priorities-based strategy for three-phase intelligent recommendation and negotiating agents for cloud services | |
| Ray et al. | Modeling and verification of service allocation policies for multi-access edge computing using probabilistic model checking |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200103 |