Precise Software Resource Allocation Method for Personalized Network Services

Abstract

The present invention belongs to the technical fields of 6G communication networks, network function virtualization, and software-defined networks, and discloses a precise software-based resource allocation method for personalized network services, mainly including the following steps: checking the types of resource requirements for personalized services; calculating the average resource requirement value for personalized services; comparing the calculated requirement value with the corresponding threshold to determine the types of service requirements; and performing precise resource allocation. Through the precise classification of personalized network services and the precise modeling and solution of the resource allocation problem, the present invention provides an optimal software-based resource allocation scheme for the personalized needs of each user in a small-scale communication network.

Description

Technical Field

[0001] The present invention belongs to the fields of 6G communication networks, network function virtualization (NFV), and software-defined networking (SDN) technologies, and specifically relates to an accurate software-based resource allocation method for personalized network services. Background Art

[0002] Currently, new applications and derivative services such as industrial Internet, intelligent manufacturing, augmented virtual reality, and the metaverse are booming. To meet the requirements of more complex service types, more stringent performance indicators, and higher-order traffic services, future communication networks need to be highly elastic and flexible. Therefore, designing a highly elastic communication network to meet the diverse and differentiated new service requirements by dynamically configuring network resources in an intelligent, flexible, and efficient manner and making full use of existing network resources is one of the challenges that future 6G communication networks need to address.

[0003] Virtualization and softwareization are important technical approaches to achieve the high elasticity of future communication networks and flexible resource allocation in communication networks, and are mainly realized through network function virtualization (NFV) and software-defined networking (SDN) technologies. NFV can fully decouple the network functions and resources of traditional dedicated network devices from the underlying general-purpose hardware, and the main function of SDN is to decouple the data and control plane of traditional dedicated network devices. The existing network resource allocation methods all perform unified resource allocation and service deployment for different services proposed by users from the perspective of network resource owners. These methods focus on meeting the goals of resource owners and cannot guarantee the personalized and differentiated resource requirements of users and ensure the personalized QoS performance indicators of users. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides an accurate software-based resource allocation method for personalized network services. This method accurately classifies the personalized and differentiated resources and function requirements of network services, and then uses optimization theory methods to provide an optimal resource allocation scheme for each personalized and differentiated service.

[0005] To achieve the above object, the present invention is realized through the following technical solutions:

[0006] The present invention is an accurate software-based resource allocation method for personalized network services. This method conducts personalized demand analysis on the virtual network service requirements proposed by users, uses its prominent requirements as additional constraints, and in advance screens out nodes that meet the resource demand conditions from the underlying 6G network to form a candidate set. It uses the convex optimization method for modeling and uses optimization software and mathematical tools to solve the convex problem to obtain the personalized optimal resource allocation plan for this network service. The specific resource allocation method includes the following steps:

[0007] Step 1: Check the types of resource requirements for the personalized network service;

[0008] Step 2: Calculate the average resource demand value of the personalized network service;

[0009] Step 3: Compare the calculated average resource demand value with a selected threshold to determine the types of personalized service requirements;

[0010] Step 4: Perform corresponding accurate software-based resource allocation for the classified personalized services.

[0011] A further improvement of the present invention lies in: In step 1, the types of resource requirements include: wired resource requirements, wireless resource requirements, and both wireless and wired resource requirements. If the resource requirements of the personalized network service only have wired resource requirements, the resource allocation and deployment of this personalized network service are carried out in the 6G core network part; if the personalized network service only has wireless resource requirements, the resource allocation and deployment of the personalized network service are carried out in the 6G access network part; if this personalized network service has both wireless and wired resource requirements, the personalized network service is carried out in the access-transmission-core network.

[0012] A further improvement of the present invention lies in: In step 2, calculating the average resource demand value of the personalized network service includes: calculating the total number of nodes |TotalElement(VN)| of the personalized network service, the average wired resource demand, the average wireless resource demand, and the average element processing transmission delay.

[0013] A further improvement of the present invention lies in: The average wired resource demand of the personalized network service is:

[0014]

[0015] Among them, a and b represent the wired nodes of the network service. CPU(a) represents the computing resources of node a, Stor(a) represents the storage resources of node a, Capa(a) represents the capacity resources of node a, Band(ab) represents the bandwidth resources of link ab, VNWiredNode represents the set of wired nodes of the network service, |VNWiredNode| represents the number of wired nodes, VNLink represents the set of links of the network service, and |VNLink| represents the number of links of this personalized service;

[0016] The average wireless resource demand of the personalized network service is:

[0017]

[0018] Among them, c represents the wireless node of this personalized network service, Spec(c) represents the spectrum resources of node c, VNWireleNode represents the set of wireless nodes of the network service, and |VNWireleNode| represents the number of wireless nodes;

[0019] The average network processing delay of the personalized network service is:

[0020]

[0021] Among them, a and b represent the nodes of the personalized network service, VNNode represents the set of nodes of the personalized network service, |VNNode| represents the total number of nodes of the personalized network service, ProDelay(a) represents the processing delay of node a, and ProDelay(ab) represents the processing delay of link ab.

[0022] A further improvement of the present invention lies in: in step 3, comparing the calculated average resource demand value with the selected threshold to determine the types of personalized service requirements, including: multi-connection network requirements, high-resource-demand network services, high-delay-demand network services, and general network services. If the total number of nodes |TotalElement(VN)| of the personalized network service is greater than the corresponding threshold, then the personalized network service belongs to multi-connection network requirements; if either the average wired resource demand AveWired(VN) or the average wireless resource demand AveWirele(VN) is greater than the corresponding resource threshold, then the personalized network service belongs to high-resource-demand network services; if the average delay |AveDelay(VN)| of the personalized network service is less than the corresponding threshold, then the personalized network service belongs to high-delay-demand network services. If none of them are satisfied, then the personalized network service is classified as a general network service.

[0023] A further improvement of the present invention lies in that: personalized network services belong to the multi-connection network requirements. The specific process of modeling personalized network services is as follows:

[0024]

[0025] Constrained by

[0026]

[0027]

[0028] Among them, X represents the corresponding relationship between service nodes and communication network nodes, and Y represents the corresponding relationship between service links and communication network links. a and b represent personalized network service nodes, A and B represent any two communication network nodes, and AB represents any communication network link. represents the corresponding relationship between personalized network service node a and communication network node A. If a is deployed on A, the value is 1, otherwise it is 0. represents the corresponding relationship between service link ab and physical link AB. If service link ab is deployed on physical link AB, the value is 1, otherwise it is 0. CPU(a) represents the computing resource of node a, Stor(a) represents the storage resource of node a, Capa(a) represents the capacity resource of node a, Band(ab) represents the bandwidth resource of service link ab, Spec(a) represents the radio spectrum resource of node a, Required is a constant referring to the minimum downlink data transmission rate, P(a) and G(a) respectively represent the transmission power and channel power gain, and σ 2 represents the Gaussian white noise power.

[0029] If the personalized network service belongs to the high-resource-demand network service, the constraints for modeling are the same as above. The only difference is the objective function:

[0030]

[0031] If the personalized network service belongs to the high-latency-demand network service, the constraints for modeling are the same as above. The only difference is the objective function:

[0032]

[0033] The beneficial effects of the present invention are as follows: By accurately classifying the personalized network service demands of users, and then modeling the network service using an optimization method for the classified personalized service, a mixed-integer non-convex programming problem is obtained. A common conversion method is used to convert this problem into a convex problem that can be directly solved by an optimization problem solver such as CVX. Through a large amount of calculations, the accurate optimal resource allocation and deployment of network service requests with personalized and differentiated demands in a highly elastic 6G network are ultimately achieved. Description of the Drawings

[0034] Figure 1 It is a schematic diagram of the present invention based on a highly elastic 6G communication network and different virtual network service demands.

[0035] Figure 2 It is a flowchart of the accurate software-defined network resource allocation method in the present invention.

[0036] Figure 3 It is a schematic diagram of the results of the success rate of network service resource allocation selected in the embodiments of the present invention.

[0037] Figure 4 It is a schematic diagram of the results of the utilization rate of the computing resources of the underlying nodes selected in the embodiments of the present invention.

[0038] Figure 5 It is a schematic diagram of the results of the utilization rate of the storage resources of the underlying nodes selected in the embodiments of the present invention. Detailed Embodiments

[0039] The following will disclose the embodiments of the present invention in diagrams. For the sake of clarity, many practical details will be described together in the following narrative. However, it should be understood that these practical details are not used to limit the present invention. That is to say, in some embodiments of the present invention, these practical details are unnecessary.

[0040] The present invention provides an accurate software-defined network resource allocation method for personalized network services, which is used in a small-scale communication network and can provide an optimal and accurate software-defined resource allocation scheme for the personalized demands of each user. As Figure 1 shown, the highly elastic 6G communication network mainly consists of three parts: the access network, the transmission network, and the core network. Directly connected to the communication core network is the Internet. For clarity, the detailed topologies and resources of the core network and the Internet are not shown in Figure 1It is drawn in detail in [reference]. In the research on software-defined resource allocation methods, the communication core network and the Internet can be abstractly represented as an undirected graph composed of a large number of terminal nodes and direct links. In the present invention, the wireless resources to be allocated are spectrum resources. In a highly flexible 6G network, spectrum resources can be software-defined and sliced into resource blocks. The wired resources (nodes and links) to be allocated are: CPU, storage, capacity, and link bandwidth. At the same time, the present invention considers the node packet processing delay and the wired link packet transmission delay as performance evaluation indicators related to QoS and QoE.

[0041] As Figure 2 shown, the present invention is an accurate software-defined resource allocation method for personalized network services, specifically including the following steps:

[0042] Step 1: Check the types of resource requirements for personalized network services.

[0043] Step 2: Calculate the average resource requirement value for the personalized network service;

[0044] Step 3: Compare the calculated average resource requirement value with a selected threshold to determine the types of personalized service requirements;

[0045] Step 4: Perform corresponding accurate software-defined resource allocation for the classified personalized services.

[0046] The following further elaborates on Steps 1-4.

[0047] In Step 1, when a personalized network service requirement is received, first check the types of resources for this network requirement.

[0048] If the resource requirements of the personalized network service only include wired resource requirements, then the resource allocation and deployment of this personalized network service are carried out in the 6G core network part; if the personalized network service only has wireless resource requirements, then the resource allocation and deployment of the personalized network service are carried out in the 6G access network part; if the personalized network service has both wireless and wired resource requirements, then the personalized network service is carried out in the access-transport-core network.

[0049] In Step 2, preliminarily process the resource requirements of the personalized network service and prepare for classifying the personalized requirements. The content to be processed includes: calculating the total number of nodes |TotalElement(VN)| of the personalized network service, the average wired resource requirement, the average wireless resource requirement, and the average element processing and transmission delay.

[0050] The average wired resource requirement is:

[0051]

[0052] Among them, a and b represent the wired nodes of the network service, CPU(a) represents the computing resources of node a, Stor(a) represents the storage resources of node a, Capa(a) represents the capacity resources of node a, Band(ab) represents the bandwidth resources of link ab, VNWiredNode represents the set of wired nodes of the network service, |VNWiredNode| represents the number of wired nodes, VNLink represents the set of links of the network service, and |VNLink| represents the number of links;

[0053] The average wireless resource demand of the personalized network service is:

[0054]

[0055] Among them, c represents the wireless node of the personalized network service, Spec(c) represents the spectrum resources of node c, VNWireleNode represents the set of wireless nodes of the network service, and |VNWireleNode| represents the number of wireless nodes;

[0056] The average network processing delay of the personalized network service is:

[0057]

[0058] Among them, a and b represent the wired nodes of the network service, VNNode represents the set of nodes of the personalized network service, |VNLink| represents the total number of links of the personalized network service, ProDelay(a) represents the processing delay of node a, and ProDelay(ab) represents the transmission delay of link ab.

[0059] In step 3, compare the four values obtained in step 2 with the resource thresholds of the 6G network. It should be noted that these thresholds can be set by oneself, selected from existing 6G research reports and white papers, or take the data of the operator as a reference by cooperating with communication operators or network service providers. Generally speaking, the operator can estimate the service threshold by measuring and calculating the indicators and scales of service demands received within a certain period of time. This threshold setting based on actual needs has reference value. If the total number of nodes of the personalized network service |TotalElement(VN)| is greater than the corresponding threshold, the personalized network service belongs to the multi-connectivity network demand; if either the average wired resource demand AveWired(VN) or the average wireless resource demand AveWirele(VN) is greater than the corresponding resource threshold, the personalized network service belongs to the high resource demand network service; if the average delay |AveDelay(VN)| of the personalized network service is less than the corresponding threshold, the personalized network service belongs to the strict delay request network service. If none of them are met, the personalized network service is classified as a general network service.

[0060] In step 4, after completing the classification of the user's personalized needs, an optimization method needs to be used to model and solve the network service. According to different types of needs, the optimization objective functions for modeling are also different. Define two types of binary variables: X and Y. X represents the correspondence between service nodes and underlying 6G network nodes, while Y represents the correspondence between service links and communication network links. To further improve the optimization efficiency and reduce unnecessary binary variables, the present invention introduces the idea of a "candidate set" of nodes. The general idea of constructing a node candidate set: Each node of each service selects in advance the physical nodes that meet the conditions from the corresponding 6G network according to its own resource needs, network function needs, and processing delay needs, and forms a candidate set. The number of elements in the candidate set of each service node is equal to the number of binary variables of that node. If the number in the candidate set is smaller, then the subsequent binary variables will be fewer.

[0061] If the personalized network service belongs to the multi-connectivity network demand, the specific modeling process of the personalized network service is as follows:

[0062]

[0063] s.t.

[0064]

[0065]

[0066] Among them, Equation (1) is the objective function, aiming to achieve precise deployment and resource allocation for all nodes of this personalized network service. Equations (2) and (3) represent binary variables. Equations (3)-(8) represent the constraint requirements for wired and wireless resources of nodes. It should be noted that Required in Equation (8) is a constant, referring to the minimum downlink data transmission rate, and this rate value is set in specific simulations. In addition, P(a) and G(a) represent transmission power and channel power gain respectively, and σ 2 represents the Gaussian white noise power. Note that the wireless transmission power P(a) can also be introduced as a continuous variable for optimization. Equations (9) and (10) are the link bandwidth requirement and the link packet transmission delay requirement respectively. Equation (11) represents the relationship between the link and the node.

[0067] If the demand for this personalized network service belongs to the high resource demand category, the constraints in the modeling are the same as above, and the only difference is the objective function:

[0068]

[0069] This objective function aims to minimize the consumption of 6G underlying physical resources, so as to free up more resource space to accommodate subsequent personalized network service demands. If the demand for this personalized network service belongs to the high latency demand category, the constraints in the modeling are the same as above, and the only difference is the objective function:

[0070]

[0071] This objective function aims to minimize the total data packet transmission delay of this network service. If the demand for this personalized network service belongs to the general category, the constraints in the modeling are the same as above, and the only difference is the objective function. For the sake of convenience, the same objective function as that of the high resource demand category is directly constructed. After completing the precise method modeling for each of the above four types of network services, it can be found that these methods all belong to the mixed integer non-convex programming problem, and theoretical analysis proves that most of such problems are NP (Non-Polynomial) hard. Therefore, the model needs to be transformed. Common transformation methods can be used. After the transformation, a dedicated programming such as CVX is used for solving. When conducting simulation work, the scale of the underlying communication network needs to be set to a medium or small scale and the number of nodes should not exceed 100.

[0072] Simulation Experiment

[0073] The present invention conducts simulation experiments to demonstrate the feasibility of the proposed method and the necessity of classifying personalized differentiated services. This application mainly selects Figure 3 the success rate of network service resource allocation as shown, Figure 4The utilization rate of the underlying node computing resources shown, such as Figure 5 The utilization rate of the underlying node storage resources shown. Other simulation results and performance metrics are not selected and plotted. Four resource allocation methods are selected for evaluation in this application: the method of classification + multi-objective function optimization, the method of no classification + multi-objective function optimization, the method of no classification + single-objective function optimization, and the method of no classification + greedy resource allocation.

[0074] Main parameter settings for the simulation experiment: The underlying network consists of 50 core nodes, 20 transmission nodes, and 20 access nodes. The connection probability between various nodes is 0.5. There are five connection links between the core-transmission and transmission-access nodes. The wired resources of each core node, such as computing and storage resources, are arbitrary integers and are randomly selected within (100, 200). The transmission nodes do not need to introduce resource attributes. The wireless resources (spectrum resources) of each access node are assumed to be software-defined into resource blocks, which are arbitrary integers and are randomly selected within (100, 200). The topological attributes of each network service link are arbitrary and the number of nodes is controlled within 10. The resource demand attributes are all arbitrary integers and are randomly selected within (1, 20).

[0075] The batch processing network service mode adopted in the simulation experiment, that is, a certain amount of network service demands are processed each time, gradually increasing from processing two network services at a time to processing nine network services at a time. Each batch is repeated 500 times to ensure the stable performance of each method eventually. The applicant will introduce the remaining detailed settings and parameter indicators, such as the selection and setting of specific thresholds, in subsequent papers.

[0076] It can be seen from the above experiments that the method of first performing personalized service classification and then performing multi-objective precise resource allocation has achieved the best performance. On the one hand, by executing the personalized service classification method, the services to be allocated can be accurately classified. Then, by adopting the precise resource allocation method and selecting appropriate optimization objectives, it can not only fully meet the needs of the personalized service but also make full use of the resources of the underlying network, leaving enough resource space for subsequent network service demands. On the other hand, selecting the corresponding objective function for the classified network services instead of just using a unified and unchanging objective function can better achieve the optimal resource allocation of personalized services. If a unified objective function is always used for resource allocation of the classified personalized services, the value of classification cannot be reflected. Therefore, in order to achieve the optimal resource allocation of personalized and differentiated services, it is necessary to both accurately classify the services and set appropriate objective functions for the classified services, so as to obtain the optimal resource allocation scheme for the personalized service.

[0077] In summary, the present invention classifies personalized network services and adopts an accurate software-based network resource allocation method based on optimization methods, enabling optimal resource allocation and deployment for network service requests with personalized and differentiated requirements in a highly elastic 6G network.

[0078] The above description is only for the implementation mode of the present invention and is not intended to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. An accurate software-based resource allocation method for personalized network services, which is used in a small-scale communication network, characterized in that: The specific steps of the precise software-based resource allocation method are as follows: Step 1: Check the types of resource requirements of the personalized network service; Step 2: Calculate the average resource requirement value of the personalized network service; Step 3: Compare the calculated average resource requirement value with the selected threshold to determine the types of personalized service requirements; Step 4: Perform corresponding precise software-based resource allocation for the classified personalized services, where: The personalized network service belongs to the multi-connection type network requirements. The specific process of modeling the personalized network service is: ， Constrained by ， ， ， ， ， ， ， ， ， ， Among them, X represents the corresponding relationship between service nodes and communication network nodes, while Y represents the corresponding relationship between service links and communication network links. a and b represent personalized network service nodes, A and B represent any two communication network nodes, and AB represents any communication network link. It represents the corresponding relationship between the personalized network service node a and the communication network node A. If a is deployed on A, the value is 1; otherwise, it is 0. It represents the corresponding relationship between the service link ab and the physical link AB. If the service link ab is deployed on the physical link AB, the value is 1; otherwise, it is 0. CPU(a) represents the computing resource of node a, Stor(a) represents the storage resource of node a, Capa(a) represents the capacity resource of node a, Band(ab) represents the bandwidth resource of link ab, Spec(a) represents the wireless spectrum resource of node a, Required is a constant referring to the minimum downlink data transmission rate, and P(a) and G(a) represent the transmission power and channel power gain respectively. It represents the Gaussian white noise power.

2. The precise software-based resource allocation method for personalized network services according to claim 1, wherein: In Step 1, the types of resource requirements include: wired resource requirements, wireless resource requirements, and both wireless and wired resource requirements. If the resource requirements of the personalized network service only have wired resource requirements, the resource allocation and deployment of this personalized network service are carried out in the 6G core network part; if the personalized network service only has wireless resource requirements, the resource allocation and deployment of the personalized network service are carried out in the 6G access network part; if this personalized network service has both wireless and wired resource requirements, the personalized network service is carried out in the access-transmission-core network.

3. The precise software-based resource allocation method for personalized network services according to claim 1, characterized in that: In Step 2, calculating the average resource requirement value of the personalized network service includes: calculating the total number of nodes |TotalElement(VN)| of the personalized network service, the average wired resource requirement, the average wireless resource requirement, and the average element processing and transmission delay.

4. The precise software-based resource allocation method for personalized network services according to claim 3, characterized in that: The average wired resource requirement of the personalized network service is: ， Where a and b represent the wired nodes of the network service, CPU(a) represents the computing resource of node a, Stor(a) represents the storage resource of node a, Capa(a) represents the capacity resource of node a, Band(ab) represents the bandwidth resource of link ab, VNWiredNode represents the set of wired nodes of the network service, |VNWiredNode| represents the number of wired nodes, VNLink represents the set of links of the network service, and |VNLink| represents the number of links; The average wireless resource requirement of the personalized network service is: Where c represents the wireless node of the network service, Spec(c) represents the wireless spectrum resource of node c, VNWireleNode represents the set of wireless nodes of the network service, and |VNWireleNode| represents the number of wireless nodes; The average network processing delay of the personalized network service is: ， Where a and b represent the nodes of the personalized network service, VNNode represents the set of nodes of the personalized network service, |VNNode| represents the total number of nodes of this personalized network service, VNLink represents the set of links of this personalized network service, |VNLink| represents the total number of links of this service, ProDelay(a) represents the processing delay of node a, and ProDelay(ab) represents the transmission delay of link ab.

5. The precise software-based resource allocation method for personalized network services according to claim 4, characterized in that: In step 3, comparing the calculated average resource demand value with the selected threshold to determine the types of personalized service requirements, including: multi-connection network requirements, high-resource-demand network services, high-latency-demand network services, and general network services. If the total number of nodes |TotalElement(VN)| of the personalized network service is greater than the corresponding threshold, the personalized network service belongs to the multi-connection network requirements; if either the average wired resource demand AveWired(VN) or the average wireless resource demand AveWirele(VN) is greater than the corresponding resource threshold, the personalized network service belongs to the high-resource-demand network services; if the average delay |AveDelay(VN)| of the personalized network service is less than the corresponding threshold, the personalized network service belongs to the high-latency-demand network services; if none of them are met, the personalized network service is classified as a general network service.

6. The precise software-based resource allocation method for personalized network services according to claim 5, characterized in that: The personalized network service belongs to the high-resource-demand network services. The specific modeling process of the personalized network service is as follows: ， Constrained by ， ， ， ， ， ， ， ， ， ， Among them, X represents the corresponding relationship between the service node and the communication network node, and Y represents the corresponding relationship between the service link and the communication network link. a and b represent personalized network service nodes, A and B represent any two communication network nodes, and AB represents any communication network link. represents the corresponding relationship between the personalized network service node a and the communication network node A. If a is deployed on A, the value is 1; otherwise, it is 0. represents the corresponding relationship between the service link ab and the physical link AB. If the service link ab is deployed on the physical link AB, the value is 1; otherwise, it is 0. CPU(a) represents the computing resource of node a, Stor(a) represents the storage resource of node a, Capa(a) represents the capacity resource of node a, Band(ab) represents the bandwidth resource of link ab, Spec(a) represents the radio spectrum resource of node a, Required is a constant referring to the minimum downlink data transmission rate, and P(a) and G(a) represent the transmission power and the channel power gain respectively. represents the Gaussian white noise power.

7. The precise software-based resource allocation method for personalized network services according to claim 5, characterized in that: The personalized network service belongs to the high-latency-demand network services. The specific modeling process of the personalized network service is as follows: ， Constrained by ， ， ， ， ， ， ， ， ， ， Among them, X represents the corresponding relationship between service nodes and communication network nodes, and Y represents the corresponding relationship between service links and communication network links. a and b represent personalized network service nodes, A and B represent any two communication network nodes, and AB represents any communication network link. represents the corresponding relationship between service node a and communication network node A. If a is deployed on A, the value is 1; otherwise, it is 0. represents the corresponding relationship between service link ab and physical link AB. If service link ab is deployed on physical link AB, the value is 1; otherwise, it is 0. CPU(a) represents the computing resource of node a, Stor(a) represents the storage resource of node a, Capa(a) represents the capacity resource of node a, Band(ab) represents the bandwidth resource of link ab, Spec(a) represents the radio spectrum resource of node a, Required is a constant referring to the minimum downlink data transmission rate, and P(a) and G(a) represent the transmission power and channel power gain respectively. represents the Gaussian white noise power.

Images