Network optimization method and server

Abstract

The invention discloses a network optimization method and a server, and relates to the technical field of network communication, and the method comprises the steps: obtaining a target network topological graph of a target network; wherein the target network is a content distribution network powered by green power. And then, according to the target network topological graph, performing problem positioning on the target network to obtain a network optimization problem. And then, solving a network optimization problem by using a gradient descent algorithm to obtain an optimal solution of the problem. And then, according to the optimal solution of the problem, determining a target request rate and a target cache allocation strategy. Wherein the target request rate and the target cache allocation strategy are used for maximizing the network utility of the target network. According to the embodiment of the invention, the effective consumption of green electricity can be realized while the network utility is maximized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, and particularly relates to a network optimization method and a server. BACKGROUND

[0002] With the continuous development of a content delivery network (CDN), network nodes in the content delivery network can serve as cache sites to store contents.

[0003] In order to optimize network performance, the same content is sent to different network nodes to quickly respond to access requests of users. However, this may cause a large amount of unnecessary content to be stored in the network nodes, and further cause energy waste of the network nodes. Therefore, how to optimize network performance while reducing energy consumption has become a problem to be solved. SUMMARY

[0004] Therefore, the present application provides a network optimization method and a server, which can maximize network utility while reducing energy consumption, and further relieve environmental burden.

[0005] In a first aspect, the present application provides a network optimization method, comprising: obtaining a target network topology graph of a target network; wherein the target network is a content delivery network powered by green electricity, and the target network topology graph is a network topology graph with virtual network nodes, and the virtual network nodes are generated based on hybrid power nodes, and the hybrid power nodes are network nodes in the target network powered by green electricity and non-green electricity in a hybrid manner; locating a problem of the target network according to the target network topology graph to obtain a network optimization problem; solving the network optimization problem by using a gradient descent algorithm to obtain an optimal solution to the problem; determining a target request rate and a target cache allocation strategy according to the optimal solution to the problem; wherein the target request rate and the target cache allocation strategy are used to maximize network utility of the target network.

[0006] In a second aspect, the present application provides a network optimization device, comprising: a topology graph obtaining module, configured to obtain a target network topology graph of a target network; wherein the target network is a content delivery network powered by green electricity, and the target network topology graph is a network topology graph with virtual network nodes, and the virtual network nodes are generated based on hybrid power nodes, and the hybrid power nodes are network nodes in the target network powered by green electricity and non-green electricity in a hybrid manner; a problem locating module, configured to locate a problem of the target network according to the target network topology graph to obtain a network optimization problem; The problem solving module is configured to solve the network optimization problem by using a gradient descent algorithm to obtain an optimal solution of the problem. The information determining module is configured to determine a target request rate and a target cache allocation strategy according to the optimal solution of the problem, wherein the target request rate and the target cache allocation strategy are used to maximize the network utility of the target network.

[0007] In a third aspect, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the method in the first aspect.

[0008] In a fourth aspect, a server is provided, which includes a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor. The processor executes the computer program to implement the method in the first aspect.

[0009] In a fifth aspect, a computer program product is provided, which stores a computer program. The computer program product is executed by a processor to implement the method in the first aspect.

[0010] The above description is only a summary of the technical solutions of the present application. In order to enable one skilled in the art to better understand the technical means of the present application, the contents of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described below.

[0011] The present application provides a network optimization method and a server. The network topology graph with virtual network nodes is used to locate the problem of the target network to obtain a network optimization problem. Then, the network optimization problem is solved to obtain an optimal solution of the problem. Then, according to the optimal solution of the problem, the request rate and the cache allocation strategy that can maximize the network utility of the target network are determined. In this way, the maximization of network utility can be realized, the situation that the user experience is ignored due to the optimization of network performance can be reduced, and the user experience is improved. Moreover, since the target network topology graph is a network topology graph with virtual network nodes, and the virtual network nodes are generated based on hybrid power supply nodes, the response speed of the target network to real-time green power fluctuations can be improved, the maximization of green power utilization can be realized, the green power is preferentially consumed, the traditional power consumption is reduced, and the environmental burden is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0012] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, those skilled in the art can obtain other drawings according to these drawings without any creative effort.

[0014] Figure 1 A schematic diagram of interaction between a user equipment and a server is shown according to an embodiment of the present application; Figure 2 A process schematic diagram of accessing a content item is shown according to an embodiment of the present application; Figure 3 A flow schematic diagram of a network optimization method is shown according to an embodiment of the present application; Figure 4 A schematic diagram of an initial network topology graph is shown according to an embodiment of the present application; Figure 5 A schematic diagram of a target network topology graph is shown according to an embodiment of the present application; Figure 6 A process schematic diagram of accessing a content item based on a target network topology graph is shown according to an embodiment of the present application; Figure 7 A flow schematic diagram of another network optimization method is shown according to an embodiment of the present application; Figure 8 A schematic diagram of a target network node caching a plurality of content items is shown according to an embodiment of the present application; Figure 9 A structural schematic diagram of a network optimization apparatus is shown according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to facilitate the description of the embodiments of the present application, some technical terms and technical means related to the embodiments of the present application and the application scenarios of the embodiments of the present application will be introduced first.

[0016] With the continuous development of Internet technology, congestion control technology and traffic engineering technology are the core supporting technologies in communication networks to improve the scalability and stability of the network. However, with the rise of the fifth generation mobile communication technology (5G for short), higher requirements are put forward for the bandwidth and delay of the network, which has an impact on the cost efficiency and scalability of centralized cloud. Therefore, in order to reduce the probability of the impact of the cost efficiency and scalability of the centralized cloud, distributed system technology represented by fog computing and edge computing can be used to reduce the bandwidth occupation and transmission delay of the 5G network.

[0017] With the development of Content Delivery Network (CDN), network nodes in the content delivery network can serve as cache sites for content storage. The CDN network is a distributed network architecture composed of edge server nodes (or edge nodes, POP points) distributed around the world, aiming to improve the efficiency of Internet content transmission by accessing cached content nearby and solving network congestion problems.

[0018] That is, in the case of receiving a content access request of a user, an edge node closest to the user and with lower load among multiple edge nodes is determined as a target edge node, and it is determined whether the target content corresponding to the content access request is cached in the target edge node. If the target edge node caches the target content corresponding to the content access request, the target content can be sent to the user equipment to realize fast response of the access request, greatly shortening the data transmission distance and data transmission time. If the target edge node does not cache the target content corresponding to the content access request, the target content can be obtained from the source server (or centralized node) to cache the target content in the target edge node. In the case where the target edge node caches the target content, the target content is sent to the user equipment to provide a basis for subsequent processing of the same access request.

[0019] It can be understood that when the CDN network is not running, in response to a content access request of any user equipment, the target content corresponding to the content access request is determined from the source server, and the target content corresponding to the content access request is sent back to the user equipment. For example, please refer to the network structure shown in (a) of Figure 1 User equipment 1-4 needs to obtain the target content corresponding to the content access request from the source server A.

[0020] When the CDN network is running, in response to a content access request of any user equipment, the server closest to the user equipment is determined, and the target content corresponding to the content access request is determined from the server, and the target content corresponding to the content access request is sent back to the user equipment. For example, please refer to Figure 1In the network structure shown in (b), for each user device, the server closest to the user device in server a can be designated as the target server. That is, the target server for user device 1 is server a, for user device 2 it is server b, for user device 3 it is server c, and for user device 4 it is server d. Then, user device 1 can retrieve the target content corresponding to its content access request from server a, user device 2 can retrieve the target content corresponding to its content access request from server b, user device 3 can retrieve the target content corresponding to its content access request from server c, and user device 4 can retrieve the target content corresponding to its content access request from server d.

[0021] The following will be based on Figure 2 Taking the transmission path as an example, this section details how a user device obtains the target content item corresponding to a content access request. The transmission path includes edge nodes v1, v2, v3, v4, and a central node S. Central node S stores all content items cached in edge nodes v1-v4; that is, all cached content items in edge nodes v1-v4 are obtained through central node S. Central node S stores content item 1, content item 2, and content item 3. Edge node v1 stores content item 1, edge node v2 stores content item 2 and content item 3, edge node v3 stores content item 1, and edge node v4 stores content item 1 and content item 3. It can be seen that the cache capacity of each edge node in the transmission path is different.

[0022] In one example, such as Figure 2 As shown, in response to content access request 1, the content access request 1 is sent to edge node v1. This content access request 1 is used to retrieve content item 1. Subsequently, upon receiving content access request 1, since edge node v1 stores the content item 1 corresponding to content access request 1, it can directly return content item 1 to the user device.

[0023] In another example, such as Figure 2 As shown, in response to content access request 2, the content access request 2 is sent to edge node v1. This content access request 2 is used to retrieve content item 2. Subsequently, upon receiving content access request 2, edge node v1, not having stored content item 2 corresponding to content access request 2, needs to send the content access request 2 to edge node v2. Then, upon receiving content access request 2, edge node v2, having stored content item 2 corresponding to content access request 2, can return content item 2 to the user device through edge node v1.

[0024] As a critical infrastructure for accelerating the distribution of static content, streaming video and / or audio on-demand and live, large software and / or game downloads, and improving website and API performance, the aforementioned CDN network can improve content loading speed, reduce network latency, and enhance user experience while simultaneously increasing network availability and reliability, alleviating server load and bandwidth pressure, and lowering costs. However, the caching capacity of network nodes and the transmission capacity of connection links in a content delivery network are limited, making it impossible to store all content on the network node closest to the user's device. Therefore, choosing a reasonable caching strategy to improve CDN network throughput is crucial. Furthermore, to optimize network performance, excessive content is cached on edge nodes for faster response to user requests. However, this can lead to network nodes storing large amounts of unnecessary content, resulting in wasted energy.

[0025] To reduce energy consumption and thus alleviate environmental burden, this application embodiment uses green electricity to power the CDN network, significantly reducing the carbon emission intensity of the data center. Green electricity refers to electricity generated using renewable energy sources (such as solar and wind power). However, during the use of green electricity, there are instances where solar power is stronger during the day and weaker at night, and wind power fluctuates at night. In other words, the spatiotemporal volatility of green electricity may not be compatible with the stable load required by the data center.

[0026] Therefore, to achieve effective utilization of green electricity and reduce energy waste caused by supply-demand mismatch, this application provides a network optimization method applicable to CDN network scenarios. The method involves obtaining a target network topology map of the target network. The target network is a content delivery network powered by green electricity, and the target network topology map is a network topology map with virtual network nodes. These virtual network nodes are generated based on hybrid power supply nodes, which are network nodes in the target network that use a mixture of green and non-green electricity for power supply. Then, based on the target network topology map, the problem of the target network is located, resulting in a network optimization problem. Next, the gradient descent algorithm is used to solve the network optimization problem, obtaining the optimal solution. Finally, based on the optimal solution, the target request rate and target cache allocation strategy are determined. The target request rate and target cache allocation strategy are used to maximize the network utility of the target network.

[0027] In this embodiment, a network optimization problem is identified by using a network topology diagram with virtual network nodes. This problem is then solved to obtain the optimal solution. Based on the optimal solution, a request rate and cache allocation strategy that maximizes the network utility of the target network are determined. This not only maximizes network utility and reduces the likelihood of neglecting user experience in network performance optimization, thus improving the user experience, but also, because the target network topology diagram is generated based on hybrid power supply nodes, it improves the target network's response speed to real-time green electricity fluctuations, thereby maximizing green electricity utilization, ensuring that green electricity is prioritized for consumption, reducing traditional electricity consumption, and ultimately reducing environmental burden.

[0028] In some examples, the execution subject of this application embodiment can be a server, which can be a single server, a server cluster, a distributed server, a centralized server, a cloud server, or a computer, etc., and is not specifically limited.

[0029] The embodiments of this application will now be described in more detail with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0030] This embodiment provides a network optimization method, such as Figure 3 As shown, the method includes: S301, Obtain the target network topology map of the target network.

[0031] The target network is a content delivery network powered by green electricity. The target network topology is a network topology with virtual network nodes. The virtual network nodes are generated based on hybrid power supply nodes, which are network nodes in the target network that use a combination of green and non-green electricity for power supply.

[0032] Specifically, the server can obtain the network structure information of the target network. This network structure information can include multiple physical network nodes and the connection links (or transmission channels) between them. Physical network nodes can be central nodes or edge nodes. Then, the server can generate an initial network topology map based on this network structure information. This initial network topology map consists of physical network nodes and connection links. The network topology map is a symmetric directed graph, and data transmitted on all links in a symmetric directed graph can propagate backward along the same path.

[0033] It should be noted that the initial network topology graph G(V, E) may include a set of edge nodes v ∈ V, a set of connection links e ∈ E, and a set of central nodes S. Each connection link in the set of connection links is invertible. For example, taking an edge node set including edge node v1 and edge node v2 (i.e., v1, v2 ∈ V), and a connection link from edge node v1 to edge node v2 (i.e., (v1, v2) ∈ E) as an example, the corresponding connection link (v2, v1) also exists.

[0034] Furthermore, all content items i transmitted in the initial network topology graph G(V,E) belong to the content directory I, and all content items i have a non-empty set S. i V, the non-empty set S i The content item i is stored but does not occupy the node's cache capacity. In this embodiment, the cache capacity of each content item i is the same, but in other embodiments, the cache capacity of each content item i can be different, and there is no specific limitation. For content items i with different cache capacities, the content item i can be split to obtain content items i with the same cache capacity. The caching strategy of content item i at the edge node v is represented as x. vi ∈{0,1}. Here, the caching strategy is used to characterize whether the edge node v caches content item i, that is, x vi =0 indicates that edge node v does not have content items i and x cached. vi =1 indicates that edge node v has content item i cached. A cache strategy set consisting of multiple cache strategies is represented as X=[x vi ] v∈V，i∈I .

[0035] For example, such as Figure 4 As shown, the initial network topology may include edge nodes v1-v7 and a central node S. Edge node v1 can establish connection links with edge nodes v4, v2, and v6; edge node v2 can establish connection links with edge nodes v4, v5, and v6; edge node v3 can establish connection links with edge nodes v5, v7, and central node S; edge node v4 can establish connection links with edge node v5; edge node v5 can establish connection links with central node S; edge node v6 can establish connection links with edge node v7; and edge node v7 can establish connection links with central node S.

[0036] Furthermore, taking the transmission path as from edge node v1 to edge node v2 to edge node v3 to central node S, and assuming that edge node v1 stores content item 1, and edge node v2 stores content item 2 and content item 3, in response to content access request 2, a content access request 2 is sent to edge node v1. This content access request 2 is used to retrieve content item 2. After receiving content access request 2, since edge node v1 does not store content item 2 corresponding to content access request 2, it needs to send content access request 2 to edge node v2. Then, upon receiving content access request 2, since edge node v2 stores content item 2 corresponding to content access request 2, it can return content item 2 to the user equipment through edge node v1.

[0037] Accordingly, after generating the initial network topology map, the architecture of the initial network topology map can be adjusted based on the green electricity information to obtain the target network topology map. Thus, since the target network topology map is constructed based on green electricity information, it ensures the timely response of the target network to real-time green electricity fluctuations, providing convenient conditions for maximizing the utilization of green electricity in the future.

[0038] The aforementioned green electricity information refers to the number of nodes that green electricity can support. The target network topology diagram may include physical network nodes and virtual network nodes. Physical network nodes can be either a first physical node or a second physical node. The first physical node is a physical network node powered entirely by green electricity. The second physical node is a physical network node that does not use green electricity. Virtual network nodes can be either a first virtual node or a second virtual node. The first virtual node is a virtual network node powered entirely by green electricity. The second virtual node is a virtual network node that does not use green electricity.

[0039] In some cases, the server can split the hybrid power supply nodes in the initial network topology diagram to obtain a first virtual node and a second virtual node. Specifically, the set of hybrid power supply nodes included in the target network topology diagram G'(V', E') is V. P In the case of any hybrid power supply node v i ∈V P , will hybrid power supply node v i Split into the first virtual node v' i_pG and the second virtual node v' i_pT Furthermore, multiple first virtual nodes v' i_pG The first set of virtual nodes is V. pG Composed of multiple second virtual nodes v' i_pT The second set of virtual nodes is V. pT .

[0040] It can be understood that the target network topology graph G'(V', E') mentioned above can include the first entity node set V. G The second set of entity nodes V T The first set of virtual nodes is V. pG And the second set of virtual nodes is V pT In other words, the set of all network nodes in the target network topology graph G'(V', E') can be V', where V' = V G ∪ V T ∪ V pG ∪ V pT Due to arbitrary hybrid power supply nodes v i ∈V P , and (v i v j Since (v') ∈ E, we can determine (v') i_pG ,v' j )∈ E' and (v' i_pT ,v' j )∈ E'. Where E' is the set of connection links in the target network topology graph G'(V',E'), and the set of caching strategies in the target network topology graph G'(V',E') is X'.

[0041] For example, such as Figure 5 As shown, the target network topology graph can include entity edge nodes v ’ 1. First virtual node v ’ 2_pG Second virtual node v ’ 2_pT Entity edge node v ’ 3- v ’ 7 and entity cluster node S. Entity edge node v ’ 1 can interact with entity edge nodes v ’ 4. First virtual node v ’ 2_pG Second virtual node v ’ 2_pT and entity edge node v ’ 6. Establish a connection link, first virtual node v ’ 2_pG Can be used with entity edge node v ’ 4. Entity edge node v ’ 3. Second virtual node v ’ 2_pT and entity edge node v ’ 6. Establish a connection link, second virtual node v ’ 2_pT Can be used with entity edge node v’ 4. Entity edge node v ’ 5. Entity edge node v ’ 3 and entity edge node v ’ 6. Establish connection links, entity edge nodes v ’ 3 can interact with entity edge nodes v ’ 5. Entity edge node v ’ 7. Establish a connection link with the entity central node S and the entity edge node v. ’ 4 can be associated with entity edge nodes v ’ 5. Establish connection links, entity edge nodes v ’ 5 can establish a connection link with the entity central node S, and the entity edge node v ’ 6 can be associated with entity edge node v ’ 7. Establish connection links, entity edge node v ’ 7 can establish a connection link with the entity centralized node S.

[0042] Furthermore, such as Figure 6 As shown, based on the transmission path, the entity edge node v ’ 1 to the first virtual node v ’ 2_pG To the second virtual node v ’ 2_pT To entity edge node v ’ 3 to entity set node S, and entity edge node v ’ 1. Stores content item 1, first virtual node v ’ 2_pG Stores content item 2 and the second virtual node v ’ 2_pT Taking content item 3 as an example, in response to content access request 2, a request is sent to entity edge node v. ’ 1. Send content access request 2. Content access request 2 is used to retrieve content item 2. Afterwards, entity edge node v ’ 1. Upon receiving content access request 2, due to the entity edge node v ’ If content item 2 corresponding to content access request 2 is not stored in node 1, then it is necessary to continue to the first virtual node v. ’ 2_pG Send content access request 2. Afterwards, the first virtual node v ’ 2_pG Upon receiving content access request 2, due to the first virtual node v ’ 2_pG If content item 2 corresponding to content access request 2 is stored, it can be accessed through the entity edge node v. ’ 1. Return content item 2 to the user's device.

[0043] S302. Based on the target network topology diagram, locate the problem in the target network to obtain the network optimization problem.

[0044] Specifically, after obtaining the target network topology map, the server can locate problems in the target network based on the target network topology map to obtain network optimization problems.

[0045] In one implementation, such as Figure 7 As shown, the process of determining the above-mentioned network optimization problem may specifically include: S3021, Input the target network topology into the network optimization model to obtain a discrete numerical optimization problem.

[0046] The network optimization model can include constraint functions and optimization functions. Constraint functions constrain the buffer capacity of network nodes and the transmission capacity of connection links. Optimization functions determine the optimization objective of the target network to maximize its network utility. In other words, by using the buffer capacity of network nodes and the transmission capacity of connection links as constraints, and combining this with the target network topology, different constraints are assigned to green electricity nodes (i.e., the first physical node and the first virtual node) and non-green electricity nodes (i.e., the second physical node and the second virtual node), providing a foundation for the target network to maximize network utility while simultaneously accommodating green electricity consumption.

[0047] Specifically, the server can input the target network topology into the constraint function to obtain constraints on the buffer capacity of each network node and the transmission capacity of each connection link. Then, the server inputs the target network topology into the optimization function to obtain the optimization objective of the target network. Finally, the server integrates the constraints on the buffer capacity of each network node, the constraints on the transmission capacity of each connection link, and the optimization objective of the target network to obtain a discrete numerical optimization problem.

[0048] In some cases, considering that the number of cached items for each edge node v ∈ V in the initial network topology G(V, E) is finite, that is, the maximum storage capacity (or cache limit) of edge node v is C. v ∈N, where N is an integer. Therefore, the constraint on the cache capacity of edge node v can be obtained by the following expression: Expression 1; Where, x vi The caching strategy for content item i at edge node v; C v This represents the maximum storage capacity of edge node v.

[0049] Furthermore, based on the constraints on the buffer capacity of edge node v in the initial network topology G(V,E), the constraints on the buffer capacity of network nodes in the target network topology G'(V',E') can be determined. Specifically, the constraints on the buffer capacity of network nodes can be obtained through the following expression: Expression 2; Where, x' vi Caching strategy for physical network nodes; C v x' is the cache capacity of the physical network node; vi_pG The caching strategy for the first virtual node; C v_pG x' is the cache capacity of the first virtual node; vi_pT The caching strategy for the second virtual node; C v_pT V represents the cache capacity of the second virtual node. G For the first set of entity nodes, V T For the second entity node set, V pG For the first set of virtual nodes and V pT This is the second set of virtual nodes.

[0050] In other cases, since the content transmission process in the initial network topology G(V,E) is driven by content access requests r ∈ R, any content access request r in the request set R can be represented as (i, p). Here, i is the requested content item, and p is the transmission path of the content access request r, which is a sequence of network nodes {p1, p2, p3, ..., p...}. kp} V, the length of the transmission path p is k p And for each network node j=1,2,…,k p -1, this network node exists (p j p j=1 Network node p ∈ E. kp For a centralized node that stores content item i, the network node p kp ∈ S i And for each transmission path, the length k ∈ [1, k p -1],v k S i In this way, by retrieving content items through transmission path p, it can be ensured that all content access requests r in the request set R can obtain the corresponding requested content item. Furthermore, the request rate of content access requests (i, p) ∈ R is represented by λ. （i，p） The set of request rates consisting of multiple request rates is represented as λ = [λ]. （i，p） ] （i，p）∈ R .

[0051] Furthermore, the content access request r = (i, p) is located at the first network node v in the transmission path p that stores the content item i. m In the case of [the previous scenario], the content item i will be transmitted back to the user equipment along the transmission path p in reverse. Therefore, the content item located at network node v in the transmission path p is [the next scenario]. m Subsequent network nodes (such as v) m+1 ) and connection links (v m v m+1 No content access request r = (i, p) will be received. That is, the transport flow on the connection link is determined by the caching policy set X and the request rate set λ. Therefore, the flow rate on connection link (b, a) when sending content item i can be obtained by the following expression three: Expression 3; Where, ρ （b，a） (λ,X) represents the flow rate on the connection link (b,a); λ （i，p） k is the request rate for content access requests (i, p); p (a) represents the position of network node a on transmission path p.

[0052] In one implementation, the constraint on the transmission capacity of the aforementioned connection link (b, a) can be obtained through the following expression: Expression 4; Among them, C ba Let C be the transmission capacity of the link (b, a), and let C be the transmission capacity of the link (b, a). ba >0 is used to characterize the bandwidth of the connection link (a, b).

[0053] Furthermore, based on the constraints of the transmission capacity of the connecting links (b, a) in the initial network topology G(V, E), the constraints of the transmission capacity of the connecting links (b, a) in the target network topology G'(V', E') can be determined.

[0054] Specifically, for any content access request r = (i, p), if the transmission path p includes network nodes powered by green electricity, that is, the transmission path p = {p1, p2, p3, p...} j , ..., p kp}, p j V p Then, in the target network topology graph G'(V', E'), the transmission path p = {p'1, p'2, p'3, p'...} j ,…,p' kp}. Entity network node p' kpFor an entity set node that stores content item i, the entity network node p' kp The content item i stored in p' is considered to be stored in p' kp_pG Therefore, the transmission path in the target network topology G'(V', E') can be p={p'1, p'2, p'3, p'...} j ,…,p' kp_pG Furthermore, in the target network topology graph G'(V', E'), the content access request r'=(i, p), the request set is R', and the request rate of content access request (i, p) ∈ R' is represented by λ. ’ （i，p） The set of request rates consisting of multiple request rates is represented as λ. ’ =[λ ’ （i，p） ] （i，p）∈ R’ .

[0055] Furthermore, in the target network topology G'(V', E') above, the transmission capacity of the connecting link (b, a) can be represented by the following expression: Expression 5; Among them, C ’ （b，a） C represents the transmission capacity of the link (b, a) in the target network topology diagram; （b，a） C represents the transmission capacity of the link (b, a) in the initial network topology diagram; ’ （b，a_pG） C represents the transmission capacity of the connecting link (b, a_pG) in the target network topology graph; ’ （b，a_pT） Let be the transmission capacity of the connecting link (b, a_pT) in the target network topology diagram; a_pG is the first virtual node mentioned above; and a_pT is the second virtual node mentioned above.

[0056] Furthermore, in the target network topology G'(V', E') above, the flow rate of the requested flow on the connecting link (b, a) can be obtained by the following expression: Expression 6; in, λ represents the flow rate of the requested flow on the connection link (b, a); ’ （i，p） k is the request rate for content access requests (i, p); p (a) represents the position of network node a on transmission path p.

[0057] In one implementation, the constraint on the transmission capacity of the connecting link (b, a) in the target network topology G'(V', E') can be obtained by the following expression: Expression 7; Among them, C ’ ba Let C be the transmission capacity of the link (b, a), and let C be the transmission capacity of the link (b, a). ’ ba >0 is used to characterize the bandwidth of the connection link (a, b).

[0058] In some cases, improving network performance such as queue latency, throughput, and transmission rate alone may not directly improve user experience, meaning it cannot guarantee a positive user experience. Therefore, this application's embodiments use the network utility of the target network as the optimization objective, aiming to maximize network utility and thereby reduce the occurrence of neglecting user experience while optimizing network performance, thus improving the user experience.

[0059] Specifically, a utility function is assigned to each content access request r'=(i,p)∈R' in the target network topology graph G'(V',E'). , to represent user satisfaction. Here, the utility function is a continuously differentiable, non-decreasing, and strictly concave function.

[0060] In one implementation, the optimization objective of the target network described above can be obtained by the following expression: Expression 8; in, The sum of network utilities (or network utilities of the target network) of all access requests (i, p) in the request set R'. The network utility of a content access request (i, p).

[0061] S3022 uses a continuous relaxation method to transform a discrete numerical optimization problem into a continuous non-convex optimization problem.

[0062] Specifically, after obtaining the discrete numerical optimization problem described above, the server can use the successive relaxation method to transform the discrete numerical optimization problem into a continuous non-convex optimization problem. The successive relaxation method (SOR) is an iterative optimization algorithm for solving linear equation systems. It introduces a relaxation factor to adjust the iteration step size, thereby accelerating the convergence speed or improving convergence.

[0063] In some cases, for each network node in the target network topology graph, the probability cache y of that network node is calculated. ’vi .in, Furthermore, the probability cache set can be... Next, for each connection link in the target network topology, the expected value of the transmission capacity constraint of the connection link is calculated to obtain the continuous constraint of the request flow on the connection link. That is, the expected value is calculated on both sides of the inequality in expression seven above to obtain the continuous constraint of the request flow on the connection link (b, a).

[0064] In one implementation, the continuity constraint of the request flow on the aforementioned connection link (b, a) can be obtained by the following expression nine: Expression Nine; in, λ represents the flow rate of the requested flow on the connection link (b, a); ’ （i，p） k is the request rate for content access requests (i, p); p (a) represents the position of network node a on transmission path p; C ’ ba Let (b, a) be the transmission capacity of the connection link (b, a).

[0065] Furthermore, for each network node in the target network topology, the expected value of the cache capacity constraint is calculated to obtain the continuous constraint conditions for the cache capacity of the network node. That is, the expected value is calculated on both sides of the inequality in Expression 2 above to obtain the continuous constraint conditions for the cache capacity of the network node.

[0066] In one implementation, the continuity constraint on the cache capacity of the aforementioned network nodes can be obtained by the following expression: Expression 10; in, A probability cache for physical network nodes; For the probability cache of the first virtual node; This is a probability cache for the second virtual node.

[0067] Next, the continuous constraints on the cache capacity of each network node in the target network topology diagram, and the continuous constraints on the request flow on each connection link in the target network topology diagram, are combined to form a constraint set D. That is, the constraint set D is { , Expression Nine, Expression Ten}.

[0068] Next, the above set of constraints and the optimization objective of the target network are integrated to obtain a continuous nonconvex optimization problem. Specifically, the continuous nonconvex optimization problem can be obtained by the following expression eleven: Expression 11; in, Let D be the network utility of the target network; D is the set of constraints mentioned above.

[0069] S3023 uses a convex relaxation method to transform a continuous non-convex optimization problem into a continuous convex optimization problem.

[0070] Specifically, after obtaining the aforementioned continuous non-convex optimization problem, the server can employ a convex relaxation method to transform it into a continuous convex optimization problem. This convex relaxation method can be generated based on an objective inequality, which can be the Goemans-Williams inequality. This addresses the issue of NP-completeness preventing solutions in polynomial time, improving the efficiency of complexity computation and ensuring the target network can make efficient decisions regarding caching strategies and request rate control.

[0071] In some cases, the mathematical description of the above Goemans and Williamson inequality is: for any Z∈[0,1] n ,definition Thus obtain Therefore, in order to use the Goemans and Williamson inequalities, the above set of constraints needs to be transformed into expressions equivalent to the Goemans and Williamson inequalities.

[0072] Specifically, for each content access request in the aforementioned request rate set, the remaining request rate for the content access request is determined based on the content access request rate and the target request rate. The target request rate is... The request rate for content access requests is The remaining request rate for content access requests is Then, based on the remaining request rate of each content access request in the request rate set. The optimization objective of the target network is updated.

[0073] In one implementation, the update process of the optimization objective of the target network can be obtained by the following expression twelve: Expression twelve; in, The updated optimization objective (i.e., the network utility of the target network); It is a utility function; For the target request rate; This represents the remaining request rate for content access requests.

[0074] Subsequently, assuming the form of the non-convex function matches the form of the objective inequality (i.e., the constraint set transformation is complete), a non-convex function inequality corresponding to the objective inequality is generated. This non-convex function inequality can include a non-convex function, an upper bound for a convex function, and a lower bound for a convex function. The non-convex function is used to determine the constraint variables corresponding to each connection link in the target network topology.

[0075] In one implementation, the above non-convex function can be characterized by the following expression thirteen: Expression Thirteen; in, It is a non-convex function, which is used to define the constraint variables corresponding to each connection link in the target network topology graph.

[0076] Furthermore, the above non-convex functional inequalities can be characterized by the following expression, number fourteen: Expression Fourteen.

[0077] Then, according to the lower bound of convex functions The set of constraints is relaxed to a set of convex constraints.

[0078] Specifically, for each connection link in the target network topology, the continuous constraints of content access requests on that connection link are input into a non-convex function to obtain the constraint variables corresponding to the connection link. Then, the constraint set is updated according to the constraint variables corresponding to each connection link in the target network topology.

[0079] In one implementation, the update process for the above constraint set can be obtained through expression 15 as shown below: Expression 15; in, A probability cache for physical network nodes; The remaining request rate for content access requests; For the target request rate; For the probability cache of the first virtual node; For the probability cache of the second virtual node; C represents the constraint variables corresponding to the connection links. ba Let (b, a) be the transmission capacity of the connection link (b, a).

[0080] Furthermore, based on the updated set of constraints and the updated optimization objective, an optimization problem equivalent to the aforementioned continuous non-convex optimization problem is obtained. Specifically, the optimization problem equivalent to the continuous non-convex optimization problem can be obtained through the following expression sixteen: Expression sixteen; in, D' represents the updated optimization objective (i.e., the network utility of the target network); D' represents the updated set of constraints.

[0081] Then, according to the lower bound of the convex function mentioned above, the updated set of constraints is relaxed into a set of convex constraints. Specifically, the set of convex constraints can be obtained through the following expression seventeen: Expression 17; in, A probability cache for physical network nodes; The remaining request rate for content access requests; For the target request rate; For the probability cache of the first virtual node; For the probability cache of the second virtual node; C ba Let (b, a) be the transmission capacity of the connection link (b, a).

[0082] Furthermore, based on the aforementioned set of convex constraints and the updated optimization objective, a continuous convex optimization problem is obtained. In other words, by replacing the updated constraint set D' in the optimization problem equivalent to the continuous non-convex optimization problem with a set of convex constraints, a convex relaxation optimization problem can be obtained, which is the continuous convex optimization problem.

[0083] In one implementation, the above continuous convex optimization problem can be obtained by the following expression eighteen: Expression 18; in, For the updated optimization objective (i.e., the network utility of the target network); D * It is a set of convex constraint conditions.

[0084] S303 uses the gradient descent algorithm to solve the network optimization problem and obtain the optimal solution.

[0085] Specifically, after obtaining the aforementioned continuous convex optimization problem, the gradient descent algorithm can be used to solve the problem and obtain the optimal solution. Gradient descent is an optimization algorithm that iteratively finds the minimum point of the loss function by moving in the opposite direction of the gradient.

[0086] In one implementation, such as Figure 7 As shown, the process of determining the optimal solution to the above problem may specifically include: S3031 uses the gradient descent algorithm to solve a continuous convex optimization problem and obtain the global optimal solution to the convex optimization problem.

[0087] Specifically, after obtaining the aforementioned continuous convex optimization problem, the gradient descent algorithm can be used to solve the continuous convex optimization problem, thereby obtaining the global optimal solution for the convex optimization problem. .

[0088] Furthermore, after obtaining the aforementioned globally optimal solution, this globally optimal solution can be... Divided into first optimal solutions Second optimal solution Among them, the first optimal solution The second optimal solution when the cache capacity is C. For a cache capacity of C ’ Solution at that time.

[0089] In one implementation, the aforementioned cache capacity C ’ This can be obtained through expression nineteen as shown below: Expression 19; Among them, C ’ ba The second optimal solution mentioned above Corresponding cache capacity; C ba The first optimal solution mentioned above The corresponding cache capacity; The target request rate.

[0090] It should be noted that, due to the aforementioned cache capacity C ’ The corresponding set of constraints D ’ C’ D * D ’ Therefore, the optimization objective corresponding to the second optimal solution is The optimization objective is less than or equal to the global optimal solution. And the optimization objective corresponding to the global optimal solution The optimization objective is less than or equal to the first optimal solution. In other words, the optimization objective corresponding to the global optimal solution. The optimization objective is better than the second optimal solution. In other words, the network utility corresponding to the globally optimal solution is better than that of a cache capacity of C. ’ The network utility of the global optimal solution at time.

[0091] S3032, based on the global optimal solution of the convex optimization problem and the gradient descent algorithm, solve the continuous non-convex optimization problem to obtain the optimal solution of the problem.

[0092] Specifically, after obtaining the global optimal solution to the aforementioned convex optimization problem, the server can use this global optimal solution as the initial parameters for a series of non-convex optimization problems. Then, using the gradient descent algorithm, based on these initial parameters, the server solves the series of non-convex optimization problems to obtain the target optimal solution for the non-convex optimization problems, which is the optimal solution (λ) to the aforementioned problem. s y s ), where λ s For optimal request rate, y s This is the optimal probability caching strategy.

[0093] In one implementation, the initial parameters are input into an optimization problem equivalent to a continuous non-convex optimization problem to obtain a local optimum. In this case, the optimization objective corresponding to the local optimum is greater than or equal to the optimization objective corresponding to the global optimum. And the optimization objective corresponding to this global optimal solution The optimization objective is greater than or equal to the second optimal solution. Then, based on this local optimal solution... Solving the above continuous nonconvex optimization problem yields the optimal solution (λ). s y s ).

[0094] S304. Based on the optimal solution to the problem, determine the target request rate and the target cache allocation strategy.

[0095] Among them, the target request rate and the target cache allocation strategy are used to maximize the network utility of the target network, that is, to maximize the total network utility of all content access requests in the target network topology.

[0096] In one implementation, an optimal cache allocation strategy is determined based on the most optimized probabilistic caching strategy and the random rounding method in the optimal solution to the above problem, and this optimal cache allocation strategy is used as the target cache allocation strategy. Furthermore, the most optimized request rate in the optimal solution to the problem is used as the target request rate. The random rounding method is a technique that introduces randomness to determine the direction of numerical rounding, the core of which is to avoid the systematic bias that may be caused by traditional rounding (such as rounding to the nearest whole number).

[0097] Specifically, the target time period is divided into T time slices. The caching strategy set X for each time slice t is... tThey are independent of each other, and the caching strategy set X t =[x t vi ] v∈V’，i∈I caching strategy x t vi ∈{0,1}. Furthermore, the caching strategy set in each time slice t satisfies the cache capacity constraint. Then, in response to the completion of caching for all content items in the target network node, for each time slice t, the cache capacity is randomly divided using a dividing line to obtain the sequence of content items stored in the target network node within time slice t, and this sequence of content items is used as the optimal cache allocation strategy x. s .

[0098] For example, such as Figure 8 As shown, the target network node's cache capacity C v Taking a cacheable content list I={1,2,3,4} as an example, with a capacity of 3, content items 1-4 are cached sequentially up to the cache capacity C. v In response to the completion of caching for all content items in the target network node, for each time slice t, the dividing line τ is used. t The cache capacity is randomly divided, and the target network node stores the sequence of content items {1,2,4} within time slice t. It can be seen that the dividing line τ... t Crossing the probability cache sequence .

[0099] It should be noted that, for ease of understanding, [the following will be used] Figure 8 The rectangle shown represents the cache capacity C of the target network node. v And divide the rectangle into C equal-sized segments. v Rows (i.e., 3 rows). Then, starting from the beginning of the first row within the rectangle, cache item 1, thus obtaining the probabilistic cache. After content item 1 is cached, content item 2 is cached at the end position of content item 1's caching, thus achieving probabilistic caching. This continues until all content items 1-4 are cached, resulting in a probabilistic cache sequence. - .

[0100] Furthermore, this embodiment provides a network optimization device, such as... Figure 9 As shown, the device includes: a topology map acquisition module 91, a problem location module 92, a problem solving module 93, and an information determination module 94.

[0101] The topology acquisition module 91 is used to acquire the target network topology map of the target network; wherein, the target network is a content delivery network powered by green electricity, and the target network topology map is a network topology map with virtual network nodes, the virtual network nodes are generated based on hybrid power supply nodes, and the hybrid power supply nodes are network nodes in the target network that use a mixture of green electricity and non-green electricity for power supply. Problem localization module 92 is used to locate problems in the target network based on the target network topology map, and obtain network optimization problems; Problem-solving module 93 is used to solve network optimization problems using the gradient descent algorithm to obtain the optimal solution to the problem; The information determination module 94 is used to determine the target request rate and the target cache allocation strategy based on the optimal solution to the problem; wherein the target request rate and the target cache allocation strategy are used to maximize the network utility of the target network.

[0102] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem localization module 92 is also used to input the target network topology map into the network optimization model to obtain a discrete numerical optimization problem; A continuous relaxation method is used to transform the discrete numerical optimization problem into a continuous non-convex optimization problem; A convex relaxation method is used to transform a continuous non-convex optimization problem into a continuous convex optimization problem; The problem-solving module 93 is also used to solve continuous convex optimization problems using the gradient descent algorithm to obtain the optimal solution to the problem.

[0103] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem localization module 92 is also used to input the target network topology map into the constraint function to obtain the constraint conditions of the buffer capacity of each network node and the constraint conditions of the transmission capacity of each connection link in the target network topology map. The target network topology is input into the optimization function to obtain the optimization objective of the target network; where the optimization objective is to maximize the network utility of the target network. By integrating the constraints on the buffer capacity of each network node, the constraints on the transmission capacity of each connection link, and the optimization objective of the target network, a discrete numerical optimization problem is obtained.

[0104] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem localization module 92 is also used to calculate the expected value of the transmission capacity constraint of each connection link in the target network topology map, so as to obtain the continuous constraint of the request flow on the connection link. For each network node in the target network topology, the expected value of the cache capacity constraint of the network node is calculated to obtain the continuous constraint of the cache capacity of the network node. The set of constraints is composed of the continuous constraints on the cache capacity of each network node in the target network topology diagram and the continuous constraints on the request flow on each connection link in the target network topology diagram. By integrating the set of constraints and the optimization objective of the target network, a continuous non-convex optimization problem is obtained.

[0105] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem location module 92 is also used to obtain the request rate set corresponding to the target network topology map; wherein, the request rate set includes the request rates of multiple content access requests in the target network topology map; For each content access request in the request rate set, determine the remaining request rate of the content access request based on the request rate of the content access request and the target request rate. The optimization target of the target network is updated based on the remaining request rate of each content access request in the request rate set. When the relational form of the non-convex function is consistent with the relational form of the objective inequality, generate the non-convex function inequality corresponding to the objective inequality; where the non-convex function inequality includes the non-convex function, the upper bound of the convex function, and the lower bound of the convex function; According to the lower bound of the convex function, relax the set of constraints into a set of convex constraints; Based on the set of convex constraints and the updated optimization objective, a continuous convex optimization problem is obtained.

[0106] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem localization module 92 is also used to input the continuous constraint conditions of the content access request on each connection link in the target network topology into a non-convex function to obtain the constraint variables corresponding to the connection link. Update the set of constraints according to the constraint variables corresponding to each connection link in the target network topology diagram; Based on the lower bound of the convex function, the updated set of constraints is relaxed into a set of convex constraints.

[0107] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the problem-solving module 93 is also used to solve continuous convex optimization problems using the gradient descent algorithm to obtain the global optimal solution of the convex optimization problem; Based on the global optimal solution of the convex optimization problem and the gradient descent algorithm, the optimal solution of the continuous non-convex optimization problem is obtained; the optimal solution includes the optimal request rate and the optimal probabilistic caching strategy.

[0108] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the information determination module 94 is also used to determine the optimal cache allocation strategy based on the optimal probability caching strategy and the random rounding method, and to use the optimal cache allocation strategy as the target cache allocation strategy. Use the optimized request rate as the target request rate.

[0109] Furthermore, in one possible implementation of this embodiment, such as Figure 9 As shown, the topology acquisition module 91 is also used to generate an initial network topology map based on the network structure information of the target network; wherein, the initial network topology map includes multiple entity network nodes; Based on the green electricity information, the initial network topology is adjusted to obtain the target network topology; the target network topology includes at least two virtual network nodes and at least one physical network node.

[0110] It should be noted that other corresponding descriptions of the functional units involved in the network optimization device provided in this embodiment can be found in the description of the network optimization method in the above embodiments, and will not be repeated here.

[0111] Based on the network optimization method shown in the above embodiments, this embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method shown in the above embodiments.

[0112] Based on the methods shown in the above embodiments, this embodiment also provides a computer program product on which a computer program is stored, and when the computer program product is executed by a processor, it implements the methods shown in the above embodiments.

[0113] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0114] Based on the method shown in the above embodiments, and ​In order to achieve the above objectives, the present application also provides a server, which includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to implement the method shown in the above embodiment.

[0115] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0116] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0117] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0118] Through the above description of the implementation methods, those skilled in the art can clearly understand that this application can be implemented using software plus necessary general-purpose hardware platforms, or it can be implemented in hardware. Compared with current related technologies, by using a network topology diagram with virtual network nodes, the target network problem is located, resulting in a network optimization problem. Then, the network optimization problem is solved to obtain the optimal solution. Subsequently, based on the optimal solution, the request rate and cache allocation strategy that maximizes the network utility of the target network are determined. This not only maximizes network utility and reduces the occurrence of neglecting user experience due to network performance optimization, thus improving the user experience, but also, because the target network topology diagram is a network topology diagram with virtual network nodes, and the virtual network nodes are generated based on hybrid power supply nodes, this improves the target network's response speed to real-time green electricity fluctuations, thereby maximizing green electricity utilization, ensuring that green electricity is prioritized for consumption, reducing traditional electricity consumption, and thus reducing environmental burden.

[0119] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0120] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A network optimization method, characterized by, The method comprises the following steps: obtaining a target network topology graph of a target network; wherein the target network is a content distribution network powered by green electricity, and the target network topology graph is a network topology graph with virtual network nodes, the virtual network nodes being generated based on hybrid power nodes, the hybrid power nodes being network nodes in the target network powered by green electricity and non-green electricity in a hybrid manner; performing problem positioning on the target network according to the target network topology graph to obtain a network optimization problem; solving the network optimization problem by using a gradient descent algorithm to obtain a problem optimal solution; determining a target request rate and a target cache allocation strategy according to the problem optimal solution; wherein the target request rate and the target cache allocation strategy are used to maximize the network utility of the target network.

2. The method of claim 1, wherein, The step of performing problem positioning on the target network according to the target network topology graph to obtain a network optimization problem comprises the following steps: inputting the target network topology graph into a network optimization model to obtain a discrete numerical optimization problem; converting the discrete numerical optimization problem into a continuous non-convex optimization problem by using a continuous relaxation method; converting the continuous non-convex optimization problem into a continuous convex optimization problem by using a convex relaxation method; The step of solving the network optimization problem by using a gradient descent algorithm to obtain a problem optimal solution comprises the following step: solving the continuous convex optimization problem by using the gradient descent algorithm to obtain the problem optimal solution.

3. The method of claim 2, wherein, In the case where the network optimization model comprises a constraint function and an optimization function, the step of inputting the target network topology graph into a network optimization model to obtain a discrete numerical optimization problem comprises the following steps: inputting the target network topology graph into the constraint function to obtain constraint conditions of cache capacities of each network node and constraint conditions of transmission capacities of each connection link in the target network topology graph; inputting the target network topology graph into the optimization function to obtain an optimization target of the target network; wherein the optimization target is to maximize the network utility of the target network; integrating the constraint conditions of cache capacities of each network node, the constraint conditions of transmission capacities of each connection link, and the optimization target of the target network to obtain the discrete numerical optimization problem.

4. The method of claim 3, wherein, The step of converting the discrete numerical optimization problem into a continuous non-convex optimization problem by using a continuous relaxation method comprises the following steps: performing expectation value calculation on the constraint conditions of transmission capacities of each connection link in the target network topology graph to obtain continuous constraint conditions of request flows on the connection link; performing expectation value calculation on the constraint conditions of cache capacities of each network node in the target network topology graph to obtain continuous constraint conditions of cache capacities of the network node; composing a constraint condition set by using the continuous constraint conditions of cache capacities of each network node in the target network topology graph and the continuous constraint conditions of request flows on each connection link in the target network topology graph. Integrate the constraint condition set and the optimization target of the target network to obtain the continuous non-convex optimization problem.

5. The method of claim 4, wherein, The continuous non-convex optimization problem is converted into a continuous convex optimization problem by using a convex relaxation method, including: Obtaining a request rate set corresponding to the target network topology graph; wherein the request rate set includes request rates of multiple content access requests in the target network topology graph; For the request rate of each content access request in the request rate set, determining the residual request rate of the content access request according to the request rate of the content access request and a target request rate; Updating the optimization target of the target network according to the residual request rate of each content access request in the request rate set; In the case that the form of the non-convex function relationship is consistent with the form of the target inequality relationship, generating a non-convex function inequality corresponding to the target inequality; wherein the non-convex function inequality includes the non-convex function, a convex function upper bound and a convex function lower bound; Relaxing the constraint condition set into a convex constraint condition set according to the convex function lower bound; Obtaining the continuous convex optimization problem according to the convex constraint condition set and the updated optimization target.

6. The method of claim 5, wherein, The constraint condition set is relaxed into a convex constraint condition set according to the convex function lower bound, including: For each connection link in the target network topology graph, inputting the continuous constraint condition of the content access request on the connection link into the non-convex function to obtain a constraint variable corresponding to the connection link; Updating the constraint condition set according to the constraint variable corresponding to each connection link in the target network topology graph; Relaxing the updated constraint condition set into the convex constraint condition set according to the convex function lower bound.

7. The method according to any one of claims 2-6, characterized in that, Solving the continuous convex optimization problem by using the gradient descent algorithm to obtain the optimal solution of the problem, including: Solving the continuous convex optimization problem by using the gradient descent algorithm to obtain the global optimal solution of the convex optimization problem; Solving the continuous non-convex optimization problem according to the global optimal solution of the convex optimization problem and the gradient descent algorithm to obtain the optimal solution of the problem; wherein the optimal solution of the problem includes an optimized request rate and an optimized probability caching strategy.

8. The method of claim 7, wherein, Determining the target request rate and the target caching allocation strategy according to the optimal solution of the problem, including: Determining an optimized caching allocation strategy according to the optimized probability caching strategy and a random rounding method, and taking the optimized caching allocation strategy as the target caching allocation strategy; Taking the optimized request rate as the target request rate.

9. The method according to any one of claims 1-6, characterized in that, Obtaining a target network topology graph of a target network, including: Generating an initial network topology graph according to network structure information of the target network; wherein the initial network topology graph includes multiple entity network nodes; According to the green electricity information, the initial network topology graph is adjusted in architecture to obtain the target network topology graph; wherein the target network topology graph comprises at least two virtual network nodes and at least one entity network node.

10. A server comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor implements the method in any one of claims 1 to 9 when executing the computer program.

Images