An edge network topology-aware microservice deployment method

Abstract

The application discloses an edge network topology-aware micro-service deployment method, aiming at the service call relationship of the edge network topology, the service call relationship is converted into a matrix form by introducing a virtual header micro-service and a call matrix, and the traffic information between nodes is determined by constructing a traffic model; the actual available bandwidth between nodes is calculated at the node layer by using a queuing model; according to the actual available bandwidth and the propagation delay between nodes, the communication delay between micro-service instances is calculated at the micro-service layer, and the average communication delay of each service is calculated at the service layer; the optimization target and the constraint condition are determined based on the communication delay, and an optimization model is established; by introducing 0-1 variables to represent whether the micro-service is deployed on the node, a micro-service deployment scheme suitable for the edge network topology is obtained. The method can optimize the micro-service deployment scheme according to the topology structure and service demand of the edge network, effectively reduce the service delay in the edge environment, and improve the user experience.

Description

Technical Field

[0001] This invention relates to the field of computer communication technology, and in particular to a microservice deployment method with edge network topology awareness. Background Technology

[0002] With the rapid development of emerging businesses such as the Industrial Internet of Things (IIoT) and virtual reality, the data scale and computing demands of network services are showing a booming growth trend. However, due to the geographical distance from users, traditional centralized cloud computing architectures are increasingly unable to meet the service quality requirements of these emerging businesses with low latency and high reliability. In order to improve the real-time performance of services and reduce data traffic to ensure reliability, emerging computing architectures such as edge computing and fog computing have been proposed and are gradually being widely used. These architectures aim to place computing and storage resources closer to the terminal, and are particularly suitable for services with low latency and high reliability requirements. However, compared with the large and centralized computing resources in centralized cloud computing architectures, edge servers have relatively limited resources in computing, storage, and communication, and are geographically dispersed. Therefore, a single edge node is often insufficient to meet the resource requirements of emerging businesses with complex tasks. Multiple distributed edge nodes are usually required to work together to meet business needs. Among the many distributed deployment scenarios, microservice architecture has attracted much attention. In microservice architecture, an application system is broken down into a series of microservices with single business functions, and multiple microservices work closely together to achieve complete system functions. Compared to traditional monolithic applications, containerized microservice instances offer faster deployment speeds and greater portability. They can be flexibly deployed on resource-constrained and heterogeneous edge devices, making them ideal for deploying services in edge environments.

[0003] While microservice architecture helps overcome device heterogeneity and allows for flexible deployment across multiple edge nodes, breaking down complete business functions into microservices makes the call relationships between these microservices exceptionally complex, and some microservices require multiple calls to achieve full business functionality. To achieve efficient microservice calls, multiple business applications can share the same type of microservice, with each type of microservice having multiple instances. However, most existing microservice calls are based on network protocols such as gRPC and RESTful APIs. Compared to traditional monolithic applications, microservice architectures have significantly higher communication requirements. Furthermore, in edge environments where nodes are geographically dispersed, the communication capabilities between nodes are heterogeneous, thus the deployment location of microservices also has a significant impact on network performance.

[0004] However, most existing research on the aforementioned issues only considers nodes with constant bandwidth, failing to adequately address the topology of edge networks and neglecting the impact of different deployment schemes on bandwidth. It has been proven that different microservice deployment schemes can severely affect the bandwidth competition between service traffic. For example, when some links carry excessive traffic, the available bandwidth for each service flow decreases significantly, leading to increased latency. Currently, there are no solutions specifically addressing the edge network topology. Summary of the Invention

[0005] The purpose of this invention is to provide a microservice deployment method that is topology-aware in edge networks. This method can optimize the microservice deployment scheme according to the topology and business needs of the edge network, effectively reduce business latency in the edge environment, and improve user experience.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] A method for deploying microservices with edge network topology awareness, the method comprising:

[0008] Step 1: For the service call relationships in the edge network topology, the service call relationships are transformed into matrix form by introducing virtual header microservices and call matrices, and the traffic information between nodes is determined by constructing a traffic model;

[0009] Step 2: Based on the obtained traffic information between nodes, a queuing model is used to calculate the actual available bandwidth between nodes at the node level;

[0010] Step 3: Based on the actual available bandwidth and propagation latency between nodes, calculate the communication latency between microservice instances at the microservice layer, and calculate the average communication latency of each service at the business layer.

[0011] Step 4: Based on the communication delay calculated in Step 3, determine the optimization objective and constraints, and establish an optimization model;

[0012] Step 5: By introducing 0-1 variables to represent whether microservices are deployed on nodes, a heuristic algorithm or reinforcement learning algorithm is used to solve the optimization model established in Step 4, and a microservice deployment scheme suitable for the edge network topology is obtained.

[0013] As can be seen from the technical solution provided by the present invention, the above method can optimize the microservice deployment scheme according to the topology and business needs of the edge network, effectively reduce the business latency in the edge environment, and improve the user experience. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the edge network topology-aware microservice deployment method provided in an embodiment of the present invention;

[0016] Figure 2 This is a schematic diagram of the business call relationship and virtual header microservice as described in an embodiment of the present invention;

[0017] Figure 3 This is a schematic diagram of a three-tier architecture for deploying microservices in an edge environment, as described in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments, and do not constitute a limitation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0019] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Contents not described in detail in the embodiments of the present invention are prior art known to those skilled in the art. Where specific conditions are not specified in the embodiments of the present invention, they shall be performed according to conventional conditions in the art or conditions recommended by the manufacturer.

[0020] like Figure 1 The diagram shows a flowchart of an edge network topology-aware microservice deployment method according to an embodiment of the present invention. The method includes:

[0021] Step 1: For the service call relationships in the edge network topology, the service call relationships are transformed into matrix form by introducing virtual header microservices and call matrices, and the traffic information between nodes is determined by constructing a traffic model;

[0022] In this step, such as Figure 2 The diagram shown illustrates the business call relationship and virtual header microservice according to an embodiment of the present invention. Node M in the diagram... i This represents the i-th type of microservice, and the numbers on the edges between nodes represent the call order between microservices. By introducing virtual header microservices and a call matrix, the business call relationship is transformed into a matrix form.

[0023] like Figure 3The diagram shown is a three-tier architecture for deploying microservices in an edge environment according to an embodiment of the present invention. Figure 3 The system comprises a business layer, a microservice layer, and an edge node layer. Specifically, at the microservice layer, a traffic model is constructed to obtain the call probability and traffic volume between microservice instances. Then, at the node layer, based on network topology and routing relationships, traffic information between any two nodes is obtained. The specific process is as follows:

[0024] Assuming there are nodes in the network that can communicate with users and forward user requests, these nodes are called access nodes. User requests always arrive through access nodes. To describe the communication process from user requests to the header microservice, where the header microservice is the microservice that the business needs to call first, a special type of microservice M0 is defined. This microservice M0 does not consume the computing resources of the node, and each access node deploys a microservice M0 as a virtual header microservice for each business.

[0025] Using F i，j Indicates from microservice M i To Microservices M j A single call; R i，j Indicates from microservice M i To Microservices M j A single response; define F k For business W k The call matrix, where Indicates the execution of business W k From Microservices M i To Microservices M j The number of times it is called; similarly, define R. k For business W k The response matrix is ​​specifically represented as:

[0026]

[0027]

[0028] Considering that there may be multiple instances of the same microservice, this invention considers using a round-robin scheduling algorithm among multiple instances. For non-virtual header microservices, when microservice M needs to be called... i At that time, each instance M i，a The probability of being visited is the same, that is:

[0029]

[0030] For the case where i equals 0, the actual value depends on the service arrival rate of each access node;

[0031] set up Indicates the execution of business W kEach time microservice M needs to be called i When selecting instance M i，a Let the probability be . Indicates each business transaction W k When invoked, instance M i，a Call instance M j，b The average number of times is then:

[0032]

[0033] Let Req i，j For microservices M i Calling microservice M j The average size of the data requested at any given time; Indicates each business transaction W k When invoked, instance M i，a To M j，b The average data size requested is:

[0034]

[0035] Similarly, let's assume Indicates each business transaction W k When invoked, instance M i，a For instance M j，b Average number of responses; Res i，j For microservices M i To Microservices M j The average size of the response data; Indicates each business transaction W k When invoked, instance M i，a For instance M j，b The average data size of the response; since each request and response is one-to-one, we have:

[0036]

[0037]

[0038] Assume any business W k The arrivals at each access node follow a Poisson distribution. Let the set of access nodes be . For access nodes Business W k The arrival parameter is If the Poisson distribution is true, then for the entire system, the business W... k Obtained by parameter λ k The Poisson distribution, wherein:

[0039]

[0040] set up Representing instance M i，a The deployed nodes, for instances of the virtual header microservice M0, in business W k During one execution, instance M 0，a The probability of being called is:

[0041]

[0042] Therefore, in executing business W k At that time, each time microservice M is called i Instance M i，a The probability of being selected is:

[0043]

[0044] For any two nodes N in the edge network p and N q Defined with N p As the source node, N q The traffic destined for the node is direct traffic Γ pq When N p and N q When the nodes are adjacent, the total traffic is defined as all data traffic between the two nodes, including direct transmissions and forwards.

[0045] Represents node N p The set of microservice instances on the edge network, for any two nodes N p and N q In executing a business transaction W k At that time, the direct flow between the two nodes is:

[0046]

[0047] Let the set of services in the edge network be . Therefore, the direct flow between the two nodes per unit time is:

[0048]

[0049] Use U p，q Indicates from node N p To node N q The routing forwarding path, where <N x N y >∈U p，q This represents a hop in the forwarding path. Since communication between microservices on the same node is transmitted via a bus, its transmission rate is much higher than the channel transmission rate between nodes. Therefore, communication on the same node can be ignored in the constructed model. When p = q, Up，q It is an empty set;

[0050] Let the set of nodes be Then node N p With N q The total flow between them is expressed as:

[0051]

[0052] Step 2: Based on the obtained traffic information between nodes, a queuing model is used to calculate the actual available bandwidth between nodes at the node level;

[0053] In this step, considering the bandwidth sharing problem when multiple services run in parallel, a queuing model is used to analyze the available bandwidth of any adjacent nodes. Assume node N... p To N q If the arrival of the sent data packets follows a Poisson distribution, and each directed link can only process one data packet at a time, then the packet sending process between nodes is regarded as an M / M / 1 queuing model.

[0054] Assuming the data packet size in the queuing model is s, Z p，q Represents node N p To N q The bandwidth of node N p Sent to N q The arrival rate and service rate of the data packets are defined as λ. p，q and μ p，q The average sending time for each data packet is calculated as follows:

[0055]

[0056] Then node N p To node N q Average available bandwidth Z′ p，q Represented as:

[0057]

[0058] in, Represents node N p With N q The total flow between them.

[0059] Step 3: Based on the actual available bandwidth and propagation latency between nodes, calculate the communication latency between microservice instances at the microservice layer, and calculate the average communication latency of each service at the business layer.

[0060] In this step, for any node N p and N q Using V p，qLet represent the propagation delay between them. Assuming full-duplex communication between nodes and neglecting the forwarding time of intermediate nodes, the total propagation delay between any two nodes is denoted as . The calculation is as follows:

[0061]

[0062] Among them, U p，q Indicates from node N p To node N q The routing and forwarding path; <N x N y >∈U p，q V represents a hop in the routing forwarding path; x，y This represents the propagation delay from node x to node y;

[0063] set up Indicates from node N p To node N q Given the minimum available throughput of each forwarding path, then the microservice instance M... i，a The request is sent to instance M j，b Average communication delay The calculation is as follows:

[0064]

[0065] Similarly, microservice instance M i，a The response is sent to instance M j，b Average communication delay The calculation is as follows:

[0066]

[0067] Then business W k Average communication latency T per execution k Represented as:

[0068]

[0069] Step 4: Based on the communication delay calculated in Step 3, determine the optimization objective and constraints, and establish an optimization model;

[0070] In this step, considering the different latency sensitivities of different services in the edge network, the average communication latency of each service is defined as the weighted sum of the average communication latency of each type of service. Let service W... k The communication delay weight is θ k The average communication latency for each service is calculated as follows:

[0071]

[0072] The optimization model can be expressed as: given a set of business operations... Microservice collection and node set In this case, by adjusting the deployment location of the microservice instance Minimize the system's average communication delay T, i.e.:

[0073]

[0074] And define the following constraints:

[0075]

[0076] Among them, constraints (21a) and (21b) mean that the total CPU and memory resource requirements of a microservice instance deployed on a single node should not exceed the total resource of the node; constraint (21c) means that the average traffic between any two adjacent nodes should not exceed the bandwidth limit.

[0077] Step 5: By introducing 0-1 variables to represent whether microservices are deployed on nodes, a heuristic algorithm or reinforcement learning algorithm is used to solve the optimization model established in Step 4, and a microservice deployment scheme suitable for the edge network topology is obtained.

[0078] In this step, considering that microservices of the same type will not have multiple instances deployed on the same node simultaneously, this approach of distributing instances across multiple nodes as much as possible can further ensure the availability of edge services. Specifically, this is achieved by using a 0-1 variable G. p，i ∈G represents the node N p Should microservices M be deployed? i Where G is a denoted by ... Matrix; For a collection of microservices, A set of nodes;

[0079] Use M p，i To represent the location at node N p Microservices M i An instance of is used to uniquely represent a microservice instance, so equation (11) is expressed as:

[0080]

[0081] Topological relationships can also be represented using matrix form, defined as follows: To represent from node N x To node N y Does the path pass through N? p To N q This jump, therefore, can be expressed as:

[0082]

[0083] Similarly, equation (14) is expressed as:

[0084]

[0085] Equations (17) and (18) are expressed as follows:

[0086]

[0087]

[0088] Equation (19) is expressed as:

[0089]

[0090] For equation (27), and These are all constants and are independent of the location where the microservice is deployed. and and Req i，j Res i，j and Both are constants; and It is an equation about G, expressed as a function of size G. The quadratic equation;

[0091] Finally, the target T is T k The weighted sum of the terms, where the weights are known, is used to form the optimization model as follows:

[0092]

[0093] The constraints are expressed as follows:

[0094]

[0095] Where f(G) and g(G) are both about G. quadratic equation;

[0096] Constraint (28a) states that the number of instances for each type of microservice must meet the requirement, therefore a total of There are several equality constraints; constraint (28b) indicates that at most one instance of the same type of microservice can be deployed on each node; constraints (28c) and (28d) represent that the total computing and memory resources of the microservices deployed on each node cannot exceed the resource limit of the corresponding node; constraint (28e) represents that the average bandwidth requirement on each link cannot exceed the bandwidth limit of that link.

[0097] In practice, when deploying microservices in edge environments, the number of microservices and nodes is relatively large. To solve this problem, heuristic algorithms (such as genetic algorithms and particle swarm optimization) or reinforcement learning methods can be used to solve the established optimization model, thereby efficiently finding a better solution in large-scale problems.

[0098] For example, taking the genetic algorithm as an example, the solution process is as follows:

[0099] 1. Using binary encoding, each chromosome represents a deployment plan, and each chromosome has... Each gene represents a type of microservice M. i The deployment scheme is as follows: 1 represents deploying this type of microservice on the corresponding node, and 0 represents not deploying it.

[0100] 2. The average communication latency T of the system, which is represented by the chromosome of each individual, is inverted and used as the fitness function for that individual.

[0101] 3. Randomly initialize the population.

[0102] 4. Calculate the fitness of each individual in the current population.

[0103] 5. Randomly select two groups of individuals from the population, and take the two individuals with the highest fitness from each group as the fathers. Randomly exchange the genes of the two fathers to produce a new generation of individuals.

[0104] 6. Perform random mutation operations on the newly generated individuals, and randomly flip the codes on random genes according to a certain probability.

[0105] 7. Repeat steps 5 and 6 until the number of the new generation population meets the requirements.

[0106] 8. Repeat steps 4-7 and record the optimal deployment plan until the maximum number of iterations is reached.

[0107] The final deployment scheme takes into account the resource requirements of microservices and the edge topology, and can reasonably select deployment nodes for each microservice instance, effectively reducing the average communication latency of various services in the edge environment and improving the user experience.

[0108] It is worth noting that the contents not described in detail in the embodiments of the present invention belong to the prior art known to those skilled in the art.

[0109] In summary, the method described in the embodiments of the present invention has the following advantages:

[0110] (1) Based on the probability model analysis of calls between microservices, it can effectively handle various complex dependencies between microservices and improve the reliability and stability of the system;

[0111] (2) This solution can fully consider the topology of the edge network, is more suitable for the scenario environment where edge devices are scattered, and can effectively reduce communication distance and network latency, and improve the response speed of edge services;

[0112] (3) By optimizing the microservice deployment scheme, the latency of edge services can be significantly reduced, the user experience quality can be improved, and the communication latency can be reduced to improve real-time performance and interactivity, enabling users to obtain responses and results faster.

[0113] (4) This solution has a number of beneficial effects in the edge computing environment under the microservice architecture, improving system performance and user experience, and promoting the application and development of edge computing.

[0114] Furthermore, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the corresponding program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0115] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims. The information disclosed in the background section is intended only to enhance the understanding of the overall background technology of the present invention and should not be construed as an admission or implication in any way that such information constitutes prior art known to those skilled in the art.

Claims

1. A microservice deployment method with edge network topology awareness, characterized in that, The method includes: Step 1: For the service call relationships in the edge network topology, the service call relationships are transformed into matrix form by introducing virtual header microservices and call matrices, and the traffic information between nodes is determined by constructing a traffic model; In step 1, specifically, the call probability and traffic volume between microservice instances are obtained at the microservice layer through the constructed traffic model. Then, at the node layer, based on the network topology and routing relationships, traffic information between any two nodes is obtained. Specifically: Assuming there are nodes in the network that can communicate with users and forward user requests, these nodes are called access nodes. User requests always arrive through access nodes. To describe the communication process from a user request to the head microservice, where the head microservice is the microservice that the business needs to call first, a special type of microservice is defined. This microservice It does not consume the computing resources of the nodes, and a microservice is deployed on each access node. This serves as the virtual header microservice for each business function; use Indicates from microservices To microservices One call; Indicates from microservices To microservices A single response; definition For business The call matrix, where Indicates execution of business From microservices To microservices The number of times it is called; similarly, define For business The response matrix is ​​specifically represented as: （1） （2） For non-virtual header microservices, when it is necessary to call the microservice... At that time, each instance The probability of being visited is the same, that is: （3） For the case where i equals 0, the actual value depends on the service arrival rate of each access node; set up Indicates execution of business Each time a microservice needs to be called When selecting an instance Let the probability be . Indicates each business transaction When invoked, the instance Calling Instance The average number of times is then: （4） set up For microservices Calling microservices The average size of the data requested at any given time; Indicates each business transaction When invoked, the instance Towards The average data size requested is: （5） Similarly, let's assume Indicates each business transaction When invoked, the instance For example The average number of responses; For microservices To microservices The average size of the response data; Indicates each business transaction When invoked, the instance For example The average data size of the response; since each request and response is one-to-one, we have: （6） （7） Assuming any business The arrivals at each access node follow a Poisson distribution. Let the set of access nodes be . For access nodes ,business The arrival parameter is If the Poisson distribution is followed, then for the system as a whole, the business... Obtain the parameter as The Poisson distribution, wherein: （8） set up Representation of instances The deployed nodes, for virtual header microservices Examples in business During one execution process, the instance The probability of being called is: （9） Therefore, in executing business At that time, each time a microservice is called , example The probability of being selected is: （10） For any two nodes in the edge network and , defined by As the source node, Traffic destined for a node is direct traffic. ;when and When the nodes are adjacent, the total traffic is defined as all data traffic between the two nodes, including direct transmissions and forwards. ; Represents a node The set of microservice instances on the edge network, for any two nodes. and In executing a business At that time, the direct flow between the two nodes is: （11） Let the set of services in the edge network be . Therefore, the direct flow between the two nodes per unit time is: （12） use Indicates from node To the node The routing forwarding path, where Indicates a hop in the forwarding path; when hour, It is an empty set; Let the set of nodes be Then the node and The total flow between them is expressed as: （13）； Step 2: Based on the obtained traffic information between nodes, a queuing model is used to calculate the actual available bandwidth between nodes at the node level; Step 3: Based on the actual available bandwidth and propagation latency between nodes, calculate the communication latency between microservice instances at the microservice layer, and calculate the average communication latency of each service at the business layer. Step 4: Based on the communication delay calculated in Step 3, determine the optimization objective and constraints, and establish an optimization model; Step 5: By introducing 0-1 variables to represent whether microservices are deployed on nodes, a heuristic algorithm or reinforcement learning algorithm is used to solve the optimization model established in Step 4, and a microservice deployment scheme suitable for the edge network topology is obtained.

2. The edge network topology-aware microservice deployment method according to claim 1, characterized in that, In step 2, considering the bandwidth sharing problem during multi-service parallel processing, a queuing model is used to analyze the available bandwidth of any adjacent nodes. It is assumed that the nodes... Towards If the arrival of sent data packets follows a Poisson distribution, and each directed link can only process one data packet at a time, then the packet sending process between nodes can be viewed as... Queuing models; Assume that the data packet size in the queuing model is... , Represents a node arrive The bandwidth, then the node Sent to The arrival rate and service rate of the data packets are defined as follows: and The average sending time for each data packet is calculated as follows: （14） So the node To the node Average available bandwidth Represented as: （15） in, Represents a node and The total flow between them.

3. The edge network topology-aware microservice deployment method according to claim 2, characterized in that, In step 3, for any node and ,use Let represent the propagation delay between them. Assuming full-duplex communication between nodes and neglecting the forwarding time of intermediate nodes, the total propagation delay between any two nodes is denoted as . The calculation is as follows: （16） in, Indicates from node To the node The routing and forwarding path; This represents a hop in the routing forwarding path; This represents the propagation delay from node x to node y; set up Indicates from node To the node The minimum available throughput of each forwarding path, then the microservice instance The request is sent to the instance. Average communication delay The calculation is as follows: （17） Similarly, microservice instances The response is sent to the instance. Average communication delay The calculation is as follows: （18） Then business Average communication latency per execution Represented as: （19）。 4. The edge network topology-aware microservice deployment method according to claim 3, characterized in that, In step 4, the average communication latency of each service is defined as the weighted sum of the average communication latency of each type of service. Let the service... The communication delay weight is The average communication latency for each service is calculated as follows: （20） The optimization model can be expressed as: given a set of business operations... Microservice collection and node set In this case, by adjusting the deployment location of the microservice instance Minimize the average communication latency of the system ,Right now: （21） And define the following constraints: ； Among them, constraints (21a) and (21b) mean that the total demand for CPU and memory resources of microservice instances deployed on a single node should not exceed the total resources of the node; constraint (21c) means that the average traffic between any two adjacent nodes should not exceed the bandwidth limit.

5. The edge network topology-aware microservice deployment method according to claim 4, characterized in that, In step 5, specifically, 0-1 variables are used. To indicate at the node Should microservices be deployed? ,in It is a size of Matrix; For a collection of microservices, A set of nodes; use To indicate the location of the node microservices An instance of is used to uniquely represent a microservice instance, so equation (11) is expressed as: （22） definition To represent from the node To the node Has the path passed through? arrive This jump, therefore, can be expressed as: （23） Similarly, equation (14) is expressed as: （24） Equations (17) and (18) are expressed as follows: （25） （26） Equation (19) is expressed as: （27） For equation (27), and These are all constants and are independent of the location where the microservice is deployed. and ,as well as , and Both are constants; and It is about The equation is expressed as a function of size . The quadratic equation; Finally, the goal yes The weighted sum of the terms, where the weights are known, is used to form the optimization model as follows: （28） The constraints are expressed as follows: ； in, and All are about of quadratic equation; Constraint (28a) states that the number of instances for each type of microservice must meet the requirement, therefore a total of There are several equality constraints; constraint (28b) indicates that at most one instance of the same type of microservice can be deployed on each node; constraints (28c) and (28d) represent that the total computing and memory resources of the microservices deployed on each node cannot exceed the resource limit of the corresponding node; constraint (28e) represents that the average bandwidth requirement on each link cannot exceed the bandwidth limit of that link.

Images