Cloud-edge elastic PON resource management network architecture, method and system based on SDN

Abstract

The invention relates to the technical field of computing power optical access networks, in particular to a cloud-side elastic PON resource management method, system and device based on an SDN (Software Defined Network). A flexible optical distribution network ODN of the network is formed by using an array waveguide grating, partition management and global control are performed on a cloud-edge elastic PON based on an SDN technology, partition fixed-point unloading is performed on user tasks by using a time delay-load cooperative unloading method and a routing strategy, and server computing resources and link spectrum resources are allocated. The problems that distributed computing nodes in an optical access network cannot sense each other, cooperative work cannot be achieved, computing resources are wasted due to the fact that computing tasks cannot be scheduled to an optimal node for computing, and the service quality cannot be guaranteed are solved. Under the computing power optical access network architecture, the end-to-end time delay demand of task unloading and the optimal network resource allocation are considered at the same time, the overall resource utilization rate of the network is further improved, and the average end-to-end time delay and the blocking rate are reduced.

Description

Technical Field

[0001] The present invention relates to computing power optical access network technology, and in particular to an SDN-based cloud-edge elastic PON resource management network architecture, method and system. Background Art

[0002] With the rapid development of the Internet, the network edge carries more and more computing-intensive and real-time applications, such as face recognition, online games, augmented reality, virtual reality, etc. The construction of cloud computing data centers allows a large number of computing tasks to be offloaded to remote cloud computing servers to alleviate computing and storage limitations. However, compared with offloading tasks directly to cloud computing data centers, the traffic of multi-access edge computing MEC can bypass the core network and cloud computing data centers, so it can effectively reduce energy consumption and meet latency requirements. At the same time, processing task requests on MEC servers at the edge of the network can also achieve protection and isolation of some sensitive data. However, MEC servers have limited computing power and poor scalability, making it difficult to meet sudden or large-scale computing offloading requests. In addition, in the traditional edge computing mode, there is a lack of unified scheduling management between edge servers, and distributed computing resources cannot be fully utilized. Therefore, it is crucial to use the high-capacity and low-latency optical fiber links of elastic optical networks to connect cloud computing data centers and various MEC servers, reasonably offload complex tasks to the network edge or cloud, and perceive the status of computing nodes in real time.

[0003] The basic idea and main functions of SDN software-defined network are separation of control and forwarding, centralized control and open interface. Separation of control and forwarding refers to the decoupling of control plane and forwarding plane, stripping the control function of traditional switches, abstracting it into a separate control plane and handing it over to SDN controller for management. Based on its centralized control architecture, SDN controller can obtain computing power information of computing nodes in computing power network in real time. When terminal device initiates computing power service request, it will schedule computing tasks to appropriate computing nodes according to its computing power demand, server load and link status, so as to improve the utilization of computing resources and user experience. In addition, SDN architecture provides new solutions for fiber optic transmission technology. Traditional routing and forwarding algorithms have low transmission efficiency and difficult maintenance. Multiple transmission protocols need to be considered during configuration. They are not adaptable to changes in network topology and environment, and it is difficult to carry out intelligent, flexible and real-time network management. The SDN controller monitors the global network topology and bandwidth in real time, makes optimal forwarding decisions based on data flow, task type, network status, and other conditions, and sends them to each network node, so that the task offloading algorithm does not need to consider complex transmission protocols when configuring. Network nodes are only responsible for data forwarding, and the network hierarchy is clear, which reduces the difficulty of routing management and configuration, making the routing strategy of task offloading more efficient.

[0004] Currently, in terms of task offloading of cloud-edge elastic PON, the mainstream methods include integer linear programming, distance priority offloading, resource priority offloading, and static best effort offloading. At the same time, in terms of routing strategies of cloud-edge elastic PON, the mainstream methods are KSP algorithm and Dijkstra algorithm. These algorithms generally only consider one factor of latency and load.

[0005] The existing computing power optical network technology has the following technical problems:

[0006] (1) In distributed computing scenarios, the computing resources of a single computing node are fixed and limited, which may lead to insufficient or wasted computing resources and make it difficult to ensure service quality.

[0007] (2) Computing nodes cannot sense each other and cannot work together, and computing tasks cannot be scheduled to the optimal computing node for calculation. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a cloud-edge elastic PON resource management method, system and device based on SDN, which improves the existing technology, improves the overall resource utilization of the network, and reduces the average end-to-end delay and blocking rate.

[0009] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: first, a cloud-edge elastic PON resource management network architecture based on SDN is provided, including: N tunable colorless optical network units ONU, multiple mobile edge computing MEC servers, N splitters, an N-periodic array waveguide grating, N multiplexers / demultiplexers, N optical line terminals OLT, 1 core router, multiple cloud servers, and optical fiber links.

[0010] The present invention also provides a cloud-edge elastic PON resource management method based on SDN, which is performed in the cloud-edge elastic PON resource management network architecture based on SDN, and includes the following steps:

[0011] S1 collects real-time information and maps it into a weighted undirected graph G (ONU s ,OLT s ,S e ,S c ,L),ONU s It is a collection of optical network units (ONUs), OLTs s is the optical line terminal OLT collection, S e is the set of edge servers, S c is the set of cloud servers, L is the set of links;

[0012] S2 calculates the selective offloading decision variable Su. When Su=1, the computing task is offloaded to the local area. When Su=0, the computing task is offloaded to the edge area or cloud area.

[0013] When S3 computing tasks are offloaded to the local area, the server with the most remaining computing resources in the local area is selected for the user; when computing tasks are offloaded to the edge area or cloud area, the latency-load coordinated offloading method is adopted to select the destination server node for the user according to the user request;

[0014] S4 determines the routing strategy for offloading computing tasks to server nodes, and uses the improved Dijkstra algorithm to calculate the k path distances L from the source node to the destination server node. p , in ascending order, starting from the first path, calculate the average spectrum resource F of each path mean , select the path with shorter distance, sufficient spectrum resources and the largest average spectrum resources.

[0015] S5 allocates and releases server node computing resources and link spectrum resources.

[0016] After S6 determines the destination server node for offloading computing tasks and the resources allocated to it for a user request, it packages the results according to the Openflow protocol and configures the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU;

[0017] S7 calculates the average end-to-end delay, blocking rate, spectrum utilization, and computing resource utilization of successful user request establishment.

[0018] Preferably, the real-time information in S1 includes: network topology, computing resources, spectrum resources, and user requests; wherein user requests u(f,r) are divided into delay-sensitive, computing-intensive, and other types, f represents the user's spectrum resource demand, and r represents the user's computing resource demand.

[0019] Preferably, the selective unloading decision variable Su described in S2 is calculated as follows:

[0020]

[0021] Among them, round refers to rounding the calculation result, P i is the proportion of remaining computing resources in the local area, K i is the end-to-end delay in the local area, P max is the maximum remaining computing resource ratio between the local area and the edge area, P min is the minimum remaining computing resource ratio between the local area and the edge area, K maxis the maximum end-to-end delay between the local area and the edge area, K min It is the minimum end-to-end delay between the local area and the edge area. S1 and S2 are two adjustable factors.

[0022] Preferably, the process of selecting a destination server for a user according to a user request described in S3 is as follows: when the user request is delay-sensitive, a distance priority strategy DFP is adopted, that is, the server with the most computing resources in the area closest to the local area will be used as the destination server to reduce the end-to-end delay of the user request; when the user request is computationally intensive, a resource priority strategy RFP is adopted, that is, the server with the most computing resources in the area with the largest proportion of remaining computing resources will be used as the destination server to maintain load balancing of computing resources in various areas; when the user request is of other types, a distance-resource coordination strategy DRP is adopted, which comprehensively considers computing resources and end-to-end delay, calculates the server importance SID of all servers in other areas outside the local area, and offloads computing tasks to the server with the largest SID;

[0023] Preferably, the server importance level SID is calculated as follows:

[0024]

[0025] Among them, r e is the computing resource of the edge server, r c is the computing resource of the cloud server, r emax is the maximum value of edge server computing resources, r emin is the minimum value of edge server computing resources, r cmax is the maximum value of the cloud server computing resources, r cmin is the minimum value of the cloud server computing resources, μ is the end-to-end delay of computing task offloading, and μ max is the maximum end-to-end delay of computing task offloading, μ min is the minimum end-to-end delay for offloading computing tasks. x and y are two binary variables. When the server is located in the edge area, x = 1, and when the server is located in the cloud area, y = 1. α and β are two adjustable factors.

[0026] Preferably, the path distance L in S4 is p , calculated as follows:

[0027]

[0028] The average spectrum resource F mean , calculated as follows:

[0029]

[0030] Among them, L iis the distance of each link on the path, f i is the spectrum resource of each link on the path, and path is the number of links on each path.

[0031] Preferably, the average end-to-end delay T in S7 is mean , blocking rate η, spectrum utilization ε, calculate resource utilization ω, which is calculated as follows:

[0032]

[0033] Among them, U represents the number of successful user requests, T u Indicates the end-to-end delay of a user request to be successfully established, U all represents the total number of user requests, f u represents the spectrum requirement for successful establishment by the user request, F represents the total number of spectrum resources, and r u It indicates the computing resource requirement for the user to successfully establish the request, and R indicates the total number of computing resources.

[0034] The present invention also provides a cloud-edge elastic PON resource management system based on SDN, the system includes an SDN controller, which is used to execute the cloud-edge elastic PON resource management method based on SDN, and the SDN controller includes the following modules:

[0035] Information perception module: used to collect real-time information;

[0036] Network mapping module: used to map the collected real-time information into a weighted undirected graph;

[0037] Task offloading module: It is used to adopt the delay-load collaborative offloading method, select the destination server for the user according to the user's request, and pass the result to the instruction issuing module;

[0038] Resource allocation module: used to allocate and release server node computing resources and link spectrum resources, and pass the results to the instruction issuing module;

[0039] Instruction issuing module: It is used to determine the destination server node for offloading computing tasks and the resources allocated to it for a user request, package the results according to the Openflow protocol, and configure the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU.

[0040] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements an SDN-based cloud-edge elastic PON resource management method as described in any one of the above contents.

[0041] The above technical solution has the following advantages or beneficial effects: the routing strategy adopts the improved Dijkstra algorithm, which not only considers the distance of the path but also the link spectrum resources, realizes the coordination of delay and load, and considers the end-to-end delay requirements and optimal computing resource allocation of computing task offloading under the computing power optical access network architecture, thereby reducing the average end-to-end delay and blocking rate. The server computing resources and link spectrum resources are allocated to further improve the overall resource utilization of the network, avoid waste of computing resources, and ensure service quality.

[0042] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present invention and should not be regarded as limiting the scope of the present invention.

[0044] Figure 1 An architecture diagram of a cloud-edge elastic PON resource management method based on SDN provided by the present invention;

[0045] Figure 2 A module diagram of a cloud-edge elastic PON resource management system based on SDN provided by the present invention;

[0046] Figure 3 A flow chart of a cloud-edge elastic PON resource management method based on SDN provided by the present invention;

[0047] Figure 4 A cloud-edge elastic PON weighted undirected graph based on SDN provided by the present invention; DETAILED DESCRIPTION

[0048] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0049] The present invention first provides a cloud-edge elastic PON resource management network architecture based on SDN, taking a large campus scenario as an example. Figure 1As shown in the figure, it is divided into four edge areas according to geographical location and function. For example, edge area 1 and edge area 4 are the basic office facilities of the office building and the factory respectively, edge area 2 is the wireless backhaul network of the entire park, and edge area 3 is the monitoring system of the entire park. Each edge area has three tunable colorless optical network units ONU, three mobile edge computing MEC servers, and there are four cloud servers in the cloud area. Four centralized optical line terminal OLT devices form an OLT pool, and an arrayed waveguide grating with a period of 4 constitutes the flexible optical distribution network ODN of the network, which is conducive to making full use of the switching function of the OLT.

[0050] The basic model of this architecture is a flexible TWDM PON (Time and Wavelength Division Multiplexed Passive Optical Network). Thanks to this, the computing tasks in this architecture do not need to go through the core network when unloading, and the switching capacity of the OLT can be fully utilized. OLT, ONU, and server all need to communicate with the centralized control plane of the SDN controller through a unified southbound interface. When a user request arrives, the SDN controller allocates resources to it according to the user request type and computing resource usage, thereby achieving real-time perception of network resources and global control of computing task unloading.

[0051] A cloud-edge elastic PON resource management method based on SDN is performed in the cloud-edge elastic PON resource management network architecture based on SDN, and the process is as follows Figure 3 As shown, the following steps are included:

[0052] S1 collects real-time information, including network topology, computing resources, spectrum resources, and user requests. User requests u(f,r) are divided into delay-sensitive, computing-intensive, and other types. f represents the user's spectrum resource demand, and r represents the user's computing resource demand.

[0053] Map the collected real-time information into a weighted undirected graph G(ONU s ,OLT s ,S e ,S c ,L),ONU s It is a collection of optical network units (ONUs), OLTs s is the optical line terminal OLT collection, S e is the set of edge servers, S c is the set of cloud servers, and L is the set of links.

[0054] The weighted undirected graph is Figure 4As shown, the solid circles represent all server nodes, the hollow circles represent the computing resources of each server, the numbers outside the brackets on each link represent the distance of the link, and the numbers inside the brackets on each link represent the link spectrum resources.

[0055] S2 calculates the selective offloading decision variable Su. When Su=1, the computing task is offloaded to the local area. When Su=0, the computing task is offloaded to the edge area or cloud area.

[0056] The calculation process of the selective unloading decision variable Su is as follows:

[0057]

[0058] Among them, round refers to rounding the calculation result, P i is the remaining computing resource ratio in the local area, K i is the end-to-end delay in the local area, P max is the maximum remaining computing resource ratio between the local area and the edge area, P min is the minimum remaining computing resource ratio between the local area and the edge area, K max is the maximum end-to-end delay between the local area and the edge area, K min It is the minimum end-to-end delay between the local area and the edge area. S1 and S2 are two adjustable factors.

[0059] When S3 computing tasks are offloaded to the local area, the server with the most remaining computing resources in the local area is selected for the user; when computing tasks are offloaded to the edge area or cloud area, the latency-load coordinated offloading method is adopted to select the destination server node for the user according to the user request;

[0060] When the user request is delay-sensitive, the distance priority strategy DFP is adopted, that is, the server with the most computing resources in the area closest to the local area will be used as the destination server to reduce the end-to-end delay of the user request; when the user request is compute-intensive, the resource priority strategy RFP is adopted, that is, the server with the most computing resources in the area with the largest proportion of remaining computing resources will be used as the destination server to maintain the load balance of computing resources in each area; when the user request is of other types, the distance-resource coordination strategy DRP is adopted, which comprehensively considers computing resources and end-to-end delay, calculates the server importance SID of all servers in other areas outside the local area, and offloads the computing task to the server with the largest SID.

[0061] The server importance level SID is calculated as follows:

[0062]

[0063] Among them, re is the computing resource of the edge server, r c is the computing resource of the cloud server, r emax is the maximum value of edge server computing resources, r emin is the minimum value of edge server computing resources, r cmax is the maximum value of the cloud server computing resources, r cmin is the minimum value of the cloud server computing resources, μ is the end-to-end delay of computing task offloading, and μ max is the maximum end-to-end delay of computing task offloading, μ min is the minimum end-to-end delay for offloading computing tasks. x and y are two binary variables. When the server is located in the edge area, x = 1, and when the server is located in the cloud area, y = 1. α and β are two adjustable factors.

[0064] S4 determines the routing strategy for offloading computing tasks to server nodes, and uses the improved Dijkstra algorithm to calculate the k path distances L from the source node to the destination server node. p The value of K depends on the network complexity. Arrange in ascending order and calculate the average spectrum resource F of each path starting from the first path. mean , select the path with shorter distance, sufficient spectrum resources and the largest average spectrum resources (that is, the remaining spectrum resources are available, and the path with the largest average spectrum resources is found under the premise of sufficient spectrum resources). This routing strategy not only considers the distance of the path but also the spectrum resources of the path, realizes the coordination of delay and load, and can reduce the blocking rate of user request establishment at the same time.

[0065] The path distance, L, is calculated as follows:

[0066]

[0067] The average spectrum resource F mean , calculated as follows:

[0068]

[0069] Among them, L i is the distance of each link on the path, f i is the spectrum resource of each link on the path, and path is the number of links on each path.

[0070] S5 allocates and releases server node computing resources and link spectrum resources.

[0071] When the computing resources of the destination server are not less than the computing resources of the user, the optimal path from the source node to the destination server node is determined through the routing strategy and spectrum resources are allocated to this path. After the computing task is transmitted to the destination server, the server computing resources are also allocated accordingly. When a service cycle ends, the used computing resources and spectrum resources are released.

[0072] When the computing resources of the destination server are less than the computing resources of the user, or the routing strategy cannot find a path that meets the conditions, or the link spectrum resources on the found path cannot meet the user's request, the user will be blocked in all three cases and will need to queue up again.

[0073] After S6 determines the destination server node for offloading computing tasks and the resources allocated to it for a user request, it packages the results according to the Openflow protocol and configures the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU;

[0074] S7 calculates the average end-to-end delay, blocking rate, spectrum utilization, and computing resource utilization of successful user request establishment.

[0075] The calculation of end-to-end delay includes: link transmission delay T l , server processing delay T s , the end-to-end delay of local area unloading is T1, the end-to-end delay of edge area unloading is T2, and the end-to-end delay of cloud area unloading is T3, which are calculated as follows:

[0076]

[0077] T1=T s ,

[0078] T2=T l +T s ,

[0079] T3=T l +T s ,

[0080] Among them, C s is the server computing speed, c is the speed of light, and l is the link transmission distance.

[0081] The average end-to-end delay T mean , blocking rate η, spectrum utilization ε, calculate resource utilization ω, which is calculated as follows:

[0082]

[0083] Among them, U represents the number of successful user requests, T uIndicates the end-to-end delay of a user request to be successfully established, U all represents the total number of user requests, f u represents the spectrum requirement for successful establishment by the user request, F represents the total number of spectrum resources, and r u It indicates the computing resource requirement for the user to successfully establish the request, and R indicates the total number of computing resources.

[0084] The present invention also provides a cloud-edge elastic PON resource management system based on SDN, the system includes an SDN controller, wherein the implementation of a cloud-edge elastic PON resource management method based on SDN relies on the mutual cooperation of five functional modules of the SDN controller, the modules include:

[0085] Information perception module: used to collect real-time information, including network topology, computing resources, spectrum resources, and user requests;

[0086] Network mapping module: used to map the collected real-time information into a weighted undirected graph;

[0087] Task offloading module: It is used to adopt the delay-load collaborative offloading method, select the destination server for the user according to the user's request, and pass the result to the instruction issuing module;

[0088] Resource allocation module: used to allocate and release server node computing resources and link spectrum resources, and pass the results to the instruction issuing module; since allocating resources to the request will affect the subsequent real-time information mapping, the resource allocation module will give feedback to the network mapping module;

[0089] Instruction issuing module: It is used to determine the destination server node for offloading computing tasks and the resources allocated to it for a user request, package the results according to the Openflow protocol, and configure the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU.

[0090] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements an SDN-based cloud-edge elastic PON resource management method as described in any one of the above contents.

[0091] The above technical solution has the following advantages or beneficial effects: The routing strategy of the present invention adopts an improved Dijkstra algorithm, which not only considers the distance of the path but also the link spectrum resources, realizes the coordination of delay and load, and can simultaneously reduce the average end-to-end delay and blocking rate. The server computing resources and link spectrum resources are allocated to further improve the overall resource utilization of the network, avoid waste of computing resources, and ensure service quality.

[0092] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A cloud-edge elastic PON resource management device based on SDN, comprising: N tunable colorless optical network units ONU, multiple mobile edge computing MEC servers, N splitters, an N-periodic array waveguide grating, N multiplexers / demultiplexers, N optical line terminals OLT, 1 core router, multiple cloud servers, and optical fiber links.

2. A cloud-edge elastic PON resource management method based on SDN, performed in a cloud-edge elastic PON resource management network architecture based on SDN as claimed in claim 1, characterized in that the steps include: S1 collects real-time information and maps it into a weighted undirected graph G (ONU s ,OLT s ,S e ,S c ,L),ONU s It is a collection of optical network units (ONUs), OLTs s is the optical line terminal OLT collection, S e is the set of edge servers, S c is the set of cloud servers, L is the set of links; S2 calculates the selective offloading decision variable Su. When Su=1, the computing task is offloaded to the local area. When Su=0, the computing task is offloaded to the edge area or cloud area. When S3 computing tasks are offloaded to the local area, the server with the most remaining computing resources in the local area is selected for the user; when computing tasks are offloaded to the edge area or cloud area, the latency-load coordinated offloading method is adopted to select the destination server node for the user according to the user request; S4 determines the routing strategy for offloading computing tasks to server nodes, and uses the improved Dijkstra algorithm to calculate the k path distances L from the source node to the destination server node. p , in ascending order, starting from the first path, calculate the average spectrum resource F of each path mean , select the path with shorter distance, sufficient spectrum resources and the largest average spectrum resources. S5 allocates and releases server node computing resources and link spectrum resources. After S6 determines the destination server node for offloading computing tasks and the resources allocated to it for a user request, it packages the results according to the Openflow protocol and configures the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU; S7 calculates the average end-to-end delay, blocking rate, spectrum utilization, and computing resource utilization of successful user request establishment.

3. A cloud-edge elastic PON resource management method based on SDN as claimed in claim 2, characterized in that: The real-time information described in S1 includes: network topology, computing resources, spectrum resources, and user requests; wherein user requests u(f,r) are divided into delay-sensitive, computing-intensive, and other types, f represents the user's spectrum resource demand, and r represents the user's computing resource demand.

4. The SDN-based cloud-edge elastic PON resource management method according to claim 2, characterized in that: The selective unloading decision variable Su described in S2 is calculated as follows: Among them, round refers to rounding the calculation result, P i is the remaining computing resource ratio in the local area, K i is the end-to-end delay in the local area, P max is the maximum remaining computing resource ratio between the local area and the edge area, P min is the minimum remaining computing resource ratio between the local area and the edge area, K max is the maximum end-to-end delay between the local area and the edge area, K min It is the minimum end-to-end delay between the local area and the edge area. S1 and S2 are two adjustable factors.

5. The SDN-based cloud-edge elastic PON resource management method according to claim 2, characterized in that: S3 describes a process of selecting a destination server for a user based on a user request as follows: when the user request is delay-sensitive, a distance priority strategy DFP is adopted, that is, the server with the most computing resources in the area closest to the local area will be used as the destination server to reduce the end-to-end delay of the user request; when the user request is compute-intensive, a resource priority strategy RFP is adopted, that is, the server with the most computing resources in the area with the largest proportion of remaining computing resources will be used as the destination server to maintain load balancing of computing resources in various areas; when the user request is of other types, a distance-resource coordination strategy DRP is adopted, which comprehensively considers computing resources and end-to-end delay, calculates the server importance SID of all servers in other areas outside the local area, and offloads computing tasks to the server with the largest SID.

6. A cloud-edge elastic PON resource management method based on SDN as claimed in claim 5, characterized in that: The server importance level SID is calculated as follows: Among them, r e is the computing resource of the edge server, r c is the computing resource of the cloud server, r emax is the maximum value of edge server computing resources, r emin is the minimum value of edge server computing resources, r cmax is the maximum value of the cloud server computing resources, r cmin is the minimum value of the cloud server computing resources, μ is the end-to-end delay of computing task offloading, and μ max is the maximum end-to-end delay of computing task offloading, μ min is the minimum end-to-end delay for offloading computing tasks. x and y are two binary variables. When the server is located in the edge area, x = 1, and when the server is located in the cloud area, y = 1. α and β are two adjustable factors.

7. The SDN-based cloud-edge elastic PON resource management method according to claim 2, characterized in that: S4 The path distance L p , calculated as follows: The average spectrum resource F mean , calculated as follows: Among them, L i is the distance of each link on the path, f i is the spectrum resource of each link on the path, and path is the number of links on each path.

8. The SDN-based cloud-edge elastic PON resource management method according to claim 2, characterized in that: S7 The average end-to-end delay T mean , blocking rate η, spectrum utilization ε, calculate resource utilization ω, which is calculated as follows: Among them, U represents the number of successful user requests, T u Indicates the end-to-end delay of a user request to be successfully established, U all represents the total number of user requests, f u represents the spectrum requirement for successful establishment by the user request, F represents the total number of spectrum resources, and r u It indicates the computing resource requirement for the user to successfully establish the request, and R indicates the total number of computing resources.

9. A cloud-edge elastic PON resource management system based on SDN, the system comprising an SDN controller, configured to execute a cloud-edge elastic PON resource management method based on SDN as claimed in claim 2, wherein the SDN controller comprises the following modules: Information perception module: used to collect real-time information; Network mapping module: used to map the collected real-time information into a weighted undirected graph; Task offloading module: It is used to adopt the delay-load collaborative offloading method, select the destination server for the user according to the user's request, and pass the result to the instruction issuing module; Resource allocation module: used to allocate and release server node computing resources and link spectrum resources, and pass the results to the instruction issuing module; Instruction issuing module: It is used to determine the destination server node for offloading computing tasks and the resources allocated to it for a user request, package the results according to the Openflow protocol, and configure the underlying infrastructure through the southbound interface of the SDN controller to specify the specific wavelengths that need to be filtered for the ONU.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, it implements the SDN-based cloud-edge elastic PON resource management method described in any one of claims 2-8.

Images