End-to-end network slice latency equalization method and system based on stochastic network calculus

Abstract

The application relates to an end-to-end network slice delay balancing method and system based on random network calculus, and belongs to the technical field of mobile communication. The method comprises the following steps: acquiring the arrival process and service process parameters of a network slice m; calculating the delay upper bound of the network slice m on different service nodes; determining the minimum resource amount required by the network slice m in the RAN domain and the CN domain and allocating corresponding network resources according to the delay upper bound; an end-to-end slice cross-domain coordinator acquires a transmission delay in real time, compares the acquired actual transmission delay with preset delay requirements; and if the actual transmission delay does not meet the delay requirements, the cross-domain coordinator re-allocates network resources according to an end-to-end delay definition delay scaling factor and the delay scaling factor. The application can effectively balance the delay requirements of various end-to-end network slices, optimize resource utilization, improve network performance and ensure efficient transmission of various businesses.

Description

Technical Field

[0001] The present invention belongs to the field of mobile communication technology and relates to an end-to-end network slicing delay balancing method and system based on random network calculation. Background Art

[0002] With the rapid development of fifth-generation mobile communications, network slicing has become a key means of achieving network flexibility and service diversification. Network slicing divides physical network resources into multiple independent logical networks, each capable of providing customized services based on specific business needs. End-to-end network slicing plays a crucial role in 5G networks, effectively meeting the diverse service demands of vertical industries. However, with the continuous growth of business demands, how to efficiently allocate network resources and ensure the transmission latency requirements of various services has become a pressing issue.

[0003] Traditional network resource allocation methods typically rely on static allocation strategies and are unable to dynamically adapt to changes in network load, easily leading to resource waste or resource shortages. When network load is lower than expected, fixed allocated resources are underutilized, resulting in resource waste. When network load is higher than expected, fixed allocated resources cannot meet demand, leading to network congestion and degraded service quality. To better address this problem, stochastic network calculus has been introduced as a mathematical analysis tool. By describing the arrival and service processes of network traffic, it can analyze and calculate performance metrics such as network latency and data backlog. This enables more accurate latency analysis and optimization in network resource allocation, thereby improving resource utilization efficiency and ensuring service quality. However, existing stochastic network calculus-based latency analysis methods mostly focus on optimization within a single domain. While these methods improve resource utilization to some extent, they have limited research on the end-to-end latency problem across multiple domains.

[0004] A network slice service-level agreement (SLA) is an agreement between an operator and a network slice customer, encompassing end-to-end performance standards and guarantee requirements. In 5G networks, end-to-end network slicing involves the coordinated operation of multiple domains, including the radio access network (RAN) and the core network (CN). Effectively allocating resources within each domain and adjusting them in real time to ensure end-to-end transmission latency requirements has become a key research topic. Existing methods lack a comprehensive solution for end-to-end latency balancing and are unable to effectively address the issues of cross-domain resource allocation and latency optimization. Summary of the Invention

[0005] In view of this, the object of the present invention is to provide an end-to-end network slicing delay balancing method and system based on random network calculation.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for end-to-end network slicing delay balancing based on stochastic network calculus, comprising the following steps:

[0008] S1: Obtain the arrival process and service process parameters of network slice m, and determine the specific parameters of the arrival envelope and service envelope through statistical analysis methods;

[0009] S2: Based on the theory of stochastic network calculus and combined with the affine service envelope, the upper bound of the latency of network slice m at different service nodes is calculated;

[0010] S3: Determine the minimum amount of resources required by network slice m in the RAN and CN domains based on the upper bound of the latency, allocate the corresponding network resources, and allocate the required minimum amount of resources to the corresponding network slice;

[0011] S4: The end-to-end slice cross-domain coordinator obtains the transmission delay in real time and compares the actual transmission delay with the preset delay requirement. If the actual transmission delay does not meet the delay requirement, the cross-domain coordinator defines a delay scaling factor based on the end-to-end delay and reallocates network resources based on the delay scaling factor. If it meets the requirement, it is continuously monitored and there is no need to adjust resource allocation.

[0012] Furthermore, in step S1, the arrival process parameters of network slice m include the average arrival rate of data packets of service m and the burst parameter of the arrival process. The system pre-statistically analyzes the arrival of service data packets to obtain the average arrival rate. The service process parameters of network slice m include the computing resource weight factor of the service service process and the burst parameter of the service process. The computing resource weight factor is obtained by calculating the ratio of the computing resources allocated by the current service node to the service to the total computing resources on the service node.

[0013] The process of obtaining the arrival process parameters of network slice m is as follows:

[0014] The service flows are kept independent and identically distributed. The process of service data packets from the service terminal to the base station BS is modeled as a Poisson process. The arrival process of service m arriving during the period [τ, t] is:

[0015]

[0016] Among them, a m (i) represents the data packet size of service m from the service terminal to the BS in the i-th time slot, in bits;

[0017] The Poisson arrival process distribution function is P[N(t)=k]=e -λt (λt) k / k! ,λ mis the average arrival rate of data packets for service m, N(t) represents the event of data packet arrival in time slot t, k represents the number of data packets that arrive in time slot t, and the moment generating function of the Poisson process is:

[0018]

[0019] Where θ represents a free parameter, θ>0;

[0020] Given a constant packet size of v, the cumulative arrival process is A(t) = N(t)v, and we get A m The moment generating function of (τ, t) is:

[0021]

[0022] Among them, E[·] represents the expected calculation, A N (τ, t) represents the Poisson arrival process of service m arriving during the period [τ, t];

[0023] For stochastic network calculus (SNC) problems, the affine upper bound is reached:

[0024]

[0025] Using affine to reach envelope parameters and Restricted Arrival Process A m (τ,t)

[0026]

[0027] When the free parameter δ is greater than 0, the moment generating function and the exponentially bounded burst model are combined with the Chernoff bound inequality to obtain the following equation:

[0028]

[0029] Among them, ε A is the probability of violation during the arrival process;

[0030] Then we get the arrival envelope burst parameters

[0031]

[0032] Furthermore, in step S1, the upper bound of the delay of network slice m is analyzed and calculated based on the random network calculus theory, including using the arrival envelope parameter and the service envelope parameter, calculating the moment generating function of the service process through the random network calculus model, and using the service cascade theorem in combination with the affine service envelope to calculate the upper bound of the delay of network slice m in different network domains; wherein,

[0033] The cumulative number of bits processed during the time slot [τ, t] is the service process S m (τ,t), which is expressed as:

[0034]

[0035] c m (i) is the time-frequency resource blocks (RBs) and computing resources allocated by the BS to service m in time slot i, that is, the cumulative number of bits that slice m can process;

[0036] The wireless channel is modeled as a Rayleigh fading channel, and the distance from the BS to the service terminal m is d m The path loss of the transmission channel is Where α is the path loss exponent, and the instantaneous signal-to-noise ratio of service m transmitted in time slot i is:

[0037]

[0038] Among them, P m Indicates the user equipment transmit power, P N represents the noise power; g is the channel gain, whose probability density function obeys the exponential distribution. The channel gain probability density function is expressed as:

[0039]

[0040] Where λ g represents the expectation of the exponential distribution, and x represents the specific value of the channel gain.

[0041] The channel gain of each time slot is independent and identically distributed, and the channel transmission service process is expressed as:

[0042]

[0043] Where B m Indicates the RB resources allocated to service m;

[0044] Its moment generating function is:

[0045]

[0046] and:

[0047]

[0048] Then the channel transmission service process moment generating function is:

[0049]

[0050] Among them, define the variable Then we get the channel transmission service envelope parameters Burst Parameters

[0051] When the service data is transmitted to the BS, the data packet is processed by the MEC server, and the MEC server is modeled as:

[0052]

[0053] The total computing power is C MEC , Φ m represents the computing resource weight factor of business m, and σ represents the business surge;

[0054] The moment generating function of MEC computing service is:

[0055]

[0056] From this we get

[0057] RAN slice domain total service process S m (τ,t) is given by and It consists of two parts, according to the service cascade theorem Its moment generating function is expressed as:

[0058]

[0059] Among them, the free parameters θ, δ> 0, the affine service envelope upper bound

[0060] Setting the affine service envelope parameters and Restriction Service Process S m (τ,t); Under the condition that the wireless transmission service process and the MEC computing service process are independent of each other, if but:

[0061]

[0062] like but:

[0063]

[0064] Get the affine service envelope parameters K is a temporary variable;

[0065] ε S Represents the probability of service violation, and then obtains the service envelope burst parameter

[0066]

[0067] Furthermore, in step S2, the burst parameters of the arrival envelope and the service envelope, and the rate of the service envelope are substituted into the data packet delay budget formula to calculate the upper bound of the delay budget for service m:

[0068]

[0069] Among them, ε m,RAN represents the delay violation probability of the RAN slice, ε m,RAN =ε A +ε s .

[0070] Furthermore, in step S3, the end-to-end network slice network resources include resource requirements in the RAN domain and the CN domain, specifically including time-frequency resources, computing resources, and link bandwidth resources. Resource allocation is optimized based on service priority while meeting latency requirements. The resource allocation process includes: S30: allocating end-to-end network slice network resources in the RAN domain and S31: allocating end-to-end network slice network resources in the CN domain.

[0071] In step S30, the upper bound of the delay budget W of RAN slice m is m Take the upper bound of the maximum latency of the RAN slice, that is, W m =τ′ m,RAN ,Calculate the minimum amount of resources required for deterministic delay transmission,RAN slice resource allocation includes the following steps:

[0072] S301: The RAN slice controller obtains the deterministic performance indicators, average packet arrival rate, and MEC server computing capacity of slice m. The deterministic performance indicators include the latency and reliability of slice m.

[0073] S302: Evenly distribute the available time-frequency resource blocks and computing resources to all RAN slices, and record the resource blocks and computing resources initially allocated to each RAN slice;

[0074] S303: sorting the slice set in descending order according to the service priority, and allocating resources to services with higher priority first;

[0075] S304: Determine whether all slice requests have been processed. If so, stop the iteration; otherwise, continue to execute S305;

[0076] S305: Calculate the number of resource blocks required based on the service transmission requirements and analyze the upper bound of the latency of the RAN slice.

[0077] S306: The upper limit of the delay is set as the delay requirement, and the required computing resources are determined;

[0078] S307: Determine whether the calculated required resources exceed the initially allocated resources. If so, repeat S305 and S306 to increase or decrease resource blocks and computing resources to ensure that the service latency requirements are met.

[0079] S307: Update the remaining time-frequency resource blocks and computing resources in the resource pool, and record the resource allocation plan for slice m.

[0080] Furthermore, in step S31, the following steps are specifically included:

[0081] The CN slice controller needs to obtain the optimal physical link of the virtual network node VNF deployed in the domain to transmit service m, and then allocate the resources required to ensure deterministic latency based on the selected link.

[0082] Among them, the KSP algorithm in graph theory is used to sort multiple physical links according to performance indicators, and several better links are selected as candidate paths;

[0083] For the CN domain, the Internet traffic distribution follows the Gaussian distribution A(t)~(μt,σ 2 t), the latency requirement τ′ of a given CN network slice m m,CN , packet transmission probability p and packet traffic arrival distribution;

[0084] Let B i is the number of packets that the virtual network node N must process per second or the link E i The number of packets that must be transmitted per second, in bits, where i∈{1,...,k}, k=N+E; let ψ i For virtual network node N or link E i The computing resources or bandwidth resources required to process or transmit a data packet;

[0085] On the premise of determining the virtual network nodes in the transmission link of slice m, the set of data packets that slice m needs to process or transmit is expressed as: Β={Β1,Β2,…,Β k}, the minimum resource amount is expressed as:

[0086]

[0087] The parameters All B i can all be represented by B1, and the optimal value of B1 is found by using the binary search algorithm;

[0088] Based on the minimum amount of resources, the CN slice controller determines the actual resource requirements initially allocated to each virtual network node on each link.

[0089] Furthermore, in step S4, a cross-domain coordinator is introduced to handle end-to-end slices related to joint resource allocation. The specific process is as follows:

[0090] When a user or enterprise has a business request, they order a slice from the network service provider (NSP) and submit a related SLA request.

[0091] After receiving the request, the NSP converts the service demand into slice demand through the cross-domain coordinator. The cross-domain coordinator is responsible for slice orchestration, deployment and maintenance, and decomposes the SLA indicators into two sub-slice SLA indicators for the RAN domain and the CN domain. The sub-domain controller deploys the sub-slices according to the SLA indicators to ensure deterministic latency transmission of the service and continuously monitors the sub-slices, and feeds back the SLA indicators to the cross-domain coordinator in real time.

[0092] Furthermore, in step S4, the end-to-end slice cross-domain coordinator obtains the transmission delay in real time and compares the actual end-to-end delay with the delay requirement, including:

[0093] S411: The end-to-end slice cross-domain coordinator obtains the transmission delay of the network slice in each domain in real time; the end-to-end slice cross-domain coordinator uses network monitoring tools and probes to collect the transmission delay of the network slice in the RAN slice and CN slice in real time;

[0094] S412: Compare the actual transmission delay obtained with the preset delay requirement to evaluate the delay performance of the service; obtain the preset transmission requirements of each service from the SLA, including the maximum tolerable delay T m and transmission reliability ε m , the transmission delays in each domain are summarized, and the end-to-end slice cross-domain coordinator calculates the end-to-end transmission delay of each network slice;

[0095] S413: For services with excessive latency, the end-to-end slice cross-domain coordinator generates warning information, including the service ID, degree of excess, and time of excess; records the excessive service information in a log and archives it for future reference. The record content includes: service ID, actual transmission delay, preset delay requirement, and degree of excess; based on the delay exceeding the standard, the end-to-end slice cross-domain coordinator adjusts resource allocation according to the delay proportional factor.

[0096] Furthermore, in step S4, if it is necessary to reallocate network resources based on the delay proportional factor, the process is as follows:

[0097] S421: Introducing SLA decomposition, setting the delay scaling factor based on the end-to-end delay through the cross-domain coordinator;

[0098] The cross-domain coordinator defines the delay domain scaling factor χ based on the E2E delay m∈(0,1) is used to determine the RAN domain and CN domain delay budgets for the slice service m, which are respectively expressed as:

[0099] τ′ m,RAN =χ m T m

[0100] τ′ m,CN =(1 - χ m )T m

[0101] The total end - to - end delay is:

[0102] τ m,E2E =τ m,RAN +τ m,CN

[0103] Set an initial value for the E2E scaling factor χ for each service m, and sort the slice set m ∈ M in descending order of service priority; m Set an initial value for the E2E scaling factor χ for each service m, and sort the slice set m ∈ M in descending order of service priority;

[0104] S422: Dynamically adjust the delay scaling factor according to the delays fed back by each sub - domain controller;

[0105] After the RAN domain and CN domain controllers obtain the domain scaling factors, use stochastic network calculus and sub - domain slice resource allocation methods to analyze the delay - bound performance and obtain the required allocated resource amount;

[0106] Each sub - domain controller performs slice configuration operations according to the allocated resources and records the actual delay; The end - to - end slice cross - domain coordinator calculates the end - to - end transmission delay of each network slice based on the sub - domain controller, and calculates the gap ΔT between the delay requirement and the actual delay m =T m -τ m,E2E ;

[0107] The cross - domain coordinator determines whether the sum of all slice delay gaps exceeds the delay - gap adjustment threshold. If so, enter step S413;

[0108] S423: Re - allocate network resources to ensure that after the delay scaling factor is adjusted, the actual transmission delay meets the preset delay requirement; The cross - domain coordinator determines whether all slice sets have been traversed. If so, record the delay domain scaling factor, otherwise enter step S424;

[0109] S424: If the actual RAN slice delay > RAN delay budget and the actual CN slice delay < CN delay budget, increase the scaling factor of the RAN domain and decrease the scaling factor of the CN domain;

[0110]

[0111] Where, represents the upper bound of the delay proportionality factor;

[0112] If the actual delay of the RAN slice < the RAN delay budget and the actual delay of the CN slice > the CN delay budget, then reduce the proportionality factor in the RAN domain and increase the proportionality factor in the CN domain;

[0113]

[0114] where, represents the lower bound of the delay proportionality factor;

[0115] S425: Continuously monitor the adjusted resource allocation scheme;

[0116] The cross-domain coordinator repeats steps S422 - S423 according to the new proportionality factor until the sum of the slice delay gaps meets the delay gap threshold.

[0117] The present invention also proposes a system applicable to the foregoing end-to-end network slice delay equalization method based on stochastic network calculus, which includes: a cross-domain coordinator, a RAN slice controller, a CN slice controller, and a monitoring module; where,

[0118] The cross-domain coordinator is used to coordinate network resources across the joint domain and perform life cycle management operations for multi-domain slices. The cross-domain coordinator is responsible for converting service requirements into slice requirements, and the cross-domain coordinator is responsible for slice orchestration, deployment, and maintenance;

[0119] The RAN slice controller is used to allocate wireless resources and MEC computing resources to perform slice operations on services in the radio access network, ensure service transmission within the RAN domain, monitor the resource usage situation within the RAN domain and optimize resource allocation by adjusting the wireless resource allocation strategy;

[0120] The CN slice controller is used to allocate computing resources of service nodes and bandwidth resources on service links, allocate resources according to the best physical links of virtual network nodes for transmitting service m, monitor the resource usage situation within the CN domain and optimize resource allocation by adjusting the allocation of computing resources and bandwidth resources.

[0121] The beneficial effects of the present invention are as follows:

[0122] By introducing a stochastic network calculus model and a cross-domain coordinator, the present invention proposes an end-to-end network slice delay equalization method based on stochastic network calculus. Compared with the traditional fixed resource allocation strategy, this method can accurately analyze the upper bound of the delay bound, dynamically adjust resource allocation, achieve end-to-end delay equalization, improve resource utilization rate, adapt to the diverse service requirements of 5G and future networks, significantly enhance the delay control accuracy, stability, and reliability of the system, and promote network intelligent management.

[0123] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0124] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0125] Figure 1 Schematic diagram of delay bound analysis for RAN slicing using stochastic network calculus;

[0126] Figure 2 Schematic diagram of end-to-end network slicing delay balancing. DETAILED DESCRIPTION

[0127] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0128] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0129] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0130] See also Figures 1 and 2 This is a method and system for end-to-end network slicing delay balancing based on stochastic network calculus, which can dynamically adjust resource allocation, achieve end-to-end delay balancing, and improve resource utilization. The method steps are as follows:

[0131] S1: Obtain the arrival process and service process parameters of network slice m, and determine the specific parameters of the arrival envelope and service envelope through statistical analysis methods;

[0132] S2: Based on the theory of stochastic network calculus and combined with the affine service envelope, the upper bound of the latency of network slice m at different service nodes is calculated;

[0133] S3: Based on the upper bound of the latency, determine the minimum amount of resources required by network slice m in the RAN and CN domains, allocate the corresponding network resources, and allocate the required minimum amount of resources to the corresponding network slice to ensure that the transmission requirements of each service are met;

[0134] S4: The end-to-end slice cross-domain coordinator obtains the transmission delay in real time, compares the actual transmission delay obtained with the preset delay requirement, and evaluates the delay performance of the network slice;

[0135] S5: If the actual transmission delay does not meet the delay requirement, the cross-domain coordinator defines a delay scaling factor based on the end-to-end delay, and reallocates network resources based on the delay scaling factor to ensure system stability and reliability.

[0136] The end-to-end network includes a BS in the RAN domain, an MEC server close to the BS, IoT terminal devices, and multiple servers in the CN domain. Each server contains one or more virtual machines, and the servers are connected by optical fibers.

[0137] Network calculus is a mathematical theory based on queuing theory, divided into deterministic and stochastic network calculus. Because stochastic network calculus considers complex random processes while guaranteeing network latency bounds within a certain probability range, it is well-suited for services with strict latency requirements while also balancing resource utilization.

[0138] In the RAN domain, the SNC calculates the upper bound of the delay and allocates time-frequency resource blocks and computing resources based on service priority and delay bound. Figure 1 Schematic diagram of delay bound analysis for RAN slicing using stochastic network calculus.

[0139] For the same service m, all data packets have the same delay requirement τ′ m,RAN , the service data packets have a first-in-first-out policy. The relationship between the RAN slice delay violation probability and the delay bound is expressed as

[0140] P[D(t)>W m]≤ε m,RAN

[0141] Where W m is the delay bound of service m, W m ≤τ′ m,RAN Indicates that the delay bound of service m in the RAN domain is lower than the delay budget on the RAN side, D(t) represents the delay that the data packet of service m will experience in time slot t, and ε m,RAN represents the delay violation probability of the RAN slice, ε m,RAN =ε A +ε s , ε A ,ε s are the violation probabilities of the arrival process and the service process, respectively.

[0142] P[D(t)>W m ]=ε m,RAN Indicates that the maximum violation probability is reached and the upper bound of the delay takes the maximum value, W m =τ′ m,RAN The minimum amount of resources required for transmitting data packets in the RAN slice domain m can be calculated at this time.

[0143] In step S1, the arrival process parameters of network slice m include the average arrival rate of data packets of service m and the burst parameter of the arrival process; the service process parameters of network slice m include the computing resource weight factor of the service service process and the burst parameter of the service process. The specific parameters of the arrival envelope and service envelope are determined by statistical analysis methods;

[0144] The service flows are kept independent and identically distributed. The process of service data packets from the service terminal to the BS is modeled as a Poisson process. The arrival process of service m arriving during the period [τ, t] is:

[0145]

[0146] Among them, a m (i) represents the data packet size of service m from the service terminal to the BS in the i-th time slot, in bits.

[0147] The Poisson arrival process distribution function is P[N(t)=k]=e -λt (λt) k / k! ,λ m is the average arrival rate of data packets for service m, N(t) represents the event of data packet arrival in time slot t, k represents the number of data packets that arrive in time slot t, and the moment generating function of the Poisson process is:

[0148]

[0149] Where, θ represents a free parameter, and θ>0.

[0150] Given a constant packet size of v, the cumulative arrival process is A(t) = N(t)v, and we get A m The moment generating function of (τ,t) is

[0151]

[0152] Among them, E[·] represents the expected calculation, A N (τ, t) represents the Poisson arrival process of service m arriving during the period [τ, t].

[0153] For the SNC problem, the affine upper bound of the envelope is reached:

[0154]

[0155] Using affine to reach envelope parameters and Restricted Arrival Process A m (τ,t)

[0156]

[0157] When the free parameter δ is greater than 0, the moment generating function and the exponentially bounded burst model are combined with the Chernoff bound inequality to obtain the following equation:

[0158]

[0159] Among them, ε A is the probability of violation during the arrival process;

[0160] Then we get the arrival envelope burst parameters

[0161]

[0162] Furthermore, the cumulative number of bits processed during the time slot [τ, t] for service m is the service process S m (τ,t), which is expressed as:

[0163]

[0164] c m (i) is the time-frequency resource block (RB) and computing resources allocated by the BS side to service m in time slot i, and the cumulative number of bits that slice m can process.

[0165] The wireless channel is modeled as a Rayleigh fading channel, and the distance from the BS to the service terminal m is d m The path loss of the transmission channel is Where α is the path loss exponent, and the instantaneous signal-to-noise ratio of service m transmitted in time slot i is:

[0166]

[0167] Among them, P m Indicates the user equipment transmit power, P N represents the noise power; g is the channel gain, whose probability density function obeys the exponential distribution. The channel gain probability density function is expressed as:

[0168]

[0169] Where λ g represents the expectation of the exponential distribution, and x represents the specific value of the channel gain.

[0170] The channel gain of each time slot is independent and identically distributed, and the channel transmission service process is expressed as:

[0171]

[0172] Where B m Indicates the RB resources allocated to service m;

[0173] Its moment generating function is:

[0174]

[0175] and:

[0176]

[0177] Then the channel transmission service process moment generating function is:

[0178]

[0179] Among them, define the variable Then we get the channel transmission service envelope parameters Burst Parameters

[0180] When the service data is transmitted to the BS, the data packet is processed by the MEC server. The MEC server is modeled as:

[0181]

[0182] The total computing power is C MEC , Φ m represents the computing resource weight factor of service m, and σ represents the service surge.

[0183] The moment generating function of MEC computing service is:

[0184]

[0185] get

[0186] RAN slice domain total service process S m (τ,t) is given by and It consists of two parts. According to the service cascade theorem, we can get Its moment generating function is expressed as:

[0187]

[0188] Free parameters θ, δ> 0, affine service envelope upper bound

[0189] Setting the affine service envelope parameters and Restriction Service Process S m (τ,t). Under the condition that the wireless transmission service process and the MEC computing service process are independent of each other, if but:

[0190]

[0191] like but:

[0192]

[0193] Get the affine service envelope parameters Further obtain the affine service envelope burst parameters:

[0194]

[0195] In step S2, the upper bound of the delay of the network slice m is calculated based on the analysis of random network calculus theory, including using the arrival envelope parameters and the service envelope parameters, calculating the moment generating function of the service process through the random network calculus model, and using the service cascade theorem in combination with the affine service envelope to calculate the upper bound of the delay of the network slice m in different network domains.

[0196] Substitute the burst parameters of the arrival envelope and service envelope, and the rate of the service envelope into the data packet delay budget formula to calculate the upper bound of the delay budget for service m:

[0197]

[0198] In step S3, the end-to-end network slice network resources include resource requirements within the RAN domain and the CN domain, specifically including time-frequency resources, computing resources, and link bandwidth resources. Based on service priority, resource allocation is optimized while meeting latency requirements. Specifically, the resource allocation process includes: S30: allocating end-to-end network slice network resources in the RAN domain and S31: allocating end-to-end network slice network resources within the CN domain.

[0199] In step S30, the upper bound of the delay budget W of RAN slice m is m Take the upper bound of the maximum latency of the RAN slice, that is, W m =τ′ m,RAN ,Calculate the minimum amount of resources required for deterministic delay transmission,RAN slice resource allocation includes the following steps:

[0200] S301: The RAN slice controller obtains the deterministic performance indicators (latency, reliability), average packet arrival rate, and MEC server computing capacity of slice m.

[0201] S302: Allocate available time-frequency resource blocks and computing resources evenly to all RAN slices, and record the resource blocks and computing resources initially allocated to each RAN slice.

[0202] S303: Sort the slice set in descending order according to the service priority, and allocate resources first to the service with higher priority.

[0203] S304: Determine whether all slice requests have been processed. If so, stop the iteration; otherwise, continue to execute S305;

[0204] S305: Calculate the number of resource blocks required based on the service transmission requirements and analyze the upper bound of the latency of the RAN slice.

[0205] S306: The upper limit of the delay is set as the delay requirement, and the required computing resources are determined;

[0206] S307: Determine whether the calculated required resources exceed the initially allocated resources. If so, repeat S305 and S306 to increase or decrease resource blocks and computing resources to ensure that the service latency requirements are met.

[0207] S308: Update the remaining time-frequency resource blocks and computing resources in the resource pool, and record the resource allocation plan for slice m.

[0208] In step S31, it includes the following steps:

[0209] In the CN domain, the CN slice controller needs to obtain the optimal physical link for the virtual network node (VNF) transporting service m deployed within the domain and then allocate the resources required to ensure deterministic latency based on the selected link. Using the KSP algorithm from graph theory, multiple physical links are ranked according to performance metrics, and the best links are selected as candidate paths.

[0210] For the CN domain, existing research analyzes the arrival of real Internet traffic and shows that the distribution of Internet traffic follows the Gaussian distribution A(t)~(μt,σ 2 t), given the latency requirement τ′ of CN network slice m m,CN , packet transmission probability p, and packet traffic arrival distribution.

[0211] In order to meet the delay requirement, let B i ,i∈{1,...,k}, is the number of packets that the virtual network node N must process per second or the number of links E i The number of packets that must be transmitted per second, in bits, where k = N + E. i For virtual network node N or link E i The computing resources or bandwidth resources required to process or transmit a data packet are different. Different services have different processing / transmission requirements.

[0212] Under the premise of determining the VNF nodes in the transmission link of slice m, in order to meet the CN slice delay constraint, the set of data packets that slice m needs to process or transmit is obtained as follows: B = {B1, B2, ..., B k}, the minimum resource amount can be expressed as:

[0213]

[0214] The parameters All B i They can all be represented by B1, and the optimal value of B1 is found by using a binary search algorithm.

[0215] Based on the minimum resource amount, the CN slice controller can determine the actual resource requirements for the initial allocation to each VNF on each link to ensure the determinism of the latency within the CN domain.

[0216] In step S4, since network slicing requires end-to-end communication service quality assurance for the entire mobile network, cross-domain orchestration and cross-domain coordination are inevitable. In order to dynamically and effectively handle end-to-end slicing related to joint resource allocation, a cross-domain coordinator is introduced. Figure 2 Schematic diagram of end-to-end network slicing delay balancing.

[0217] When a user or enterprise requests a service, they can order a slice from a network service provider (NSP) and submit a related SLA request. Upon receiving the request, the NSP converts the service demand into a slice requirement through a cross-domain coordinator. The cross-domain coordinator is responsible for slice orchestration, deployment, and maintenance, breaking down the SLA into sub-slice SLAs for the RAN and CN domains. The sub-domain controller deploys the sub-slices based on the SLAs to ensure deterministic latency transmission of the service, continuously monitors the sub-slices, and provides real-time feedback to the cross-domain coordinator.

[0218] The end-to-end slice cross-domain coordinator in this embodiment obtains transmission delay in real time and compares the actual end-to-end delay with the delay requirement, including:

[0219] S411: The end-to-end slice cross-domain coordinator obtains the transmission delay of the network slice in each domain in real time; the end-to-end slice cross-domain coordinator collects the transmission delay of the network slice in the RAN slice and the CN slice in real time through network monitoring tools and probes.

[0220] S412: Compare the actual transmission delay obtained with the preset delay requirement to evaluate the delay performance of the service; obtain the preset transmission requirements of each service from the SLA, including the maximum tolerable delay T m and transmission reliability ε m The transmission delay in each domain is aggregated, and the end-to-end slice cross-domain coordinator calculates the end-to-end transmission delay of each network slice.

[0221] S413: For services with excessive latency, early warnings and records are issued to facilitate subsequent resource adjustments. For services with excessive latency, the end-to-end slice cross-domain coordinator generates early warning information, including detailed data such as the service ID, degree of excess, and time of excess. The detailed information about the exceeding service is recorded in a log and archived for future reference. The log includes the service ID, actual transmission latency, preset latency requirements, and degree of excess. Based on the excessive latency, the end-to-end slice cross-domain coordinator adjusts resource allocation according to the latency scaling factor.

[0222] In this embodiment, if the end-to-end slice cross-domain coordinator needs to adjust resource allocation based on the delay scale factor, the process includes:

[0223] S421: Introducing SLA decomposition, setting the delay scaling factor based on the end-to-end delay through the cross-domain coordinator;

[0224] The cross-domain coordinator defines the delay domain scaling factor χ based on the E2E delay m ∈(0,1), which is used to determine the RAN domain and CN domain delay budgets of slice service m, respectively expressed as:

[0225] τ′ m,RAN =χm T m

[0226] τ′ m,CN =(1 - χ m )T m

[0227] The total end - to - end delay is:

[0228] τ m,E2E =τ m,RAN +τ m,CN <**********><**********>Initially, set the E2E scaling factor χ m of each service m to 0.5, and sort the slice set m ∈ M in descending order of service priority.

[0230] S422: Dynamically adjust the delay scaling factor according to the delays fed back by each sub - domain controller;

[0231] After the RAN domain and CN domain controllers obtain the domain scaling factors, use stochastic network calculus and sub - domain slice resource allocation methods to analyze the delay - bound performance and obtain the required allocated resource amount.

[0232] Each sub - domain controller performs slice configuration operations according to the allocated resources and records the actual delay. The end - to - end slice cross - domain coordinator calculates the end - to - end transmission delay of each network slice based on the sub - domain controller, and calculates the gap ΔT m =T m -τ m,E2E .

[0233] The cross - domain coordinator determines whether the sum of all slice delay gaps exceeds the delay - gap adjustment threshold. If so, enter step S413.

[0234] S423: Re - allocate network resources to ensure that after the delay scaling factor is adjusted, the actual transmission delay meets the preset delay requirement; the cross - domain coordinator determines whether all slice sets have been traversed. If so, record the delay domain scaling factor, otherwise enter step S424.

[0235] S424: If the actual delay of the RAN slice > RAN delay budget and the actual delay of the CN slice < CN delay budget, increase the scaling factor of the RAN domain and decrease the scaling factor of the CN domain.

[0236]

[0237] [[ID=SS6]]Among them, represents the upper bound of the delay scaling factor.

[0238] If the actual RAN slice latency < RAN latency budget and the actual CN slice latency > CN latency budget, then reduce the scaling factor in the RAN domain and increase the scaling factor in the CN domain.

[0239]

[0240] Wherein, represents the lower bound of the latency scaling factor.

[0241] S425: Continuously monitor the adjusted resource allocation scheme to ensure the stability and reliability of the system.

[0242] The cross-domain coordinator repeats steps S422 - S423 according to the new scaling factor until the sum of the slice latency gaps meets the latency gap threshold.

[0243] On the other hand, the present invention provides a system applicable to the aforementioned end-to-end network slice latency equalization method based on stochastic network calculus, which includes: a cross-domain coordinator, a RAN slice controller, a CN slice controller, and a monitoring module;

[0244] The cross-domain coordinator is used to coordinate network resources across the joint domain and perform life cycle management operations for multi-domain slices. The cross-domain coordinator is responsible for converting service requirements into slice requirements, and is responsible for slice orchestration, deployment, and maintenance;

[0245] The RAN slice controller is used to allocate wireless resources and MEC computing resources to perform slice operations on services in the radio access network, ensure service transmission within the RAN domain, monitor resource usage within the RAN domain and optimize resource allocation, and ensure that latency requirements within the RAN domain are met by adjusting the wireless resource allocation strategy;

[0246] The CN slice controller is used to allocate computing resources of service nodes and bandwidth resources on service links, allocate resources according to the best physical link of the VNF for transmitting service m, ensure transmission latency requirements within the CN domain, monitor resource usage within the CN domain and optimize resource allocation, and ensure that latency within the CN domain meets the requirements by adjusting the allocation of computing resources and bandwidth resources.

[0247] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for end-to-end network slicing delay balancing based on stochastic network calculus, characterized by: It includes the following steps: S1: Obtain the arrival process and service process parameters of network slice m, and determine the specific parameters of the arrival envelope and service envelope through statistical analysis methods; S2: Based on the theory of stochastic network calculus and combined with the affine service envelope, the upper bound of the latency of network slice m at different service nodes is calculated; S3: Determine the minimum amount of resources required by network slice m in the RAN and CN domains based on the upper bound of the latency, allocate the corresponding network resources, and allocate the required minimum amount of resources to the corresponding network slice; S4: The end-to-end slice cross-domain coordinator obtains the transmission delay in real time and compares the actual transmission delay with the preset delay requirement; If the actual transmission delay does not meet the delay requirement, the cross-domain coordinator defines a delay scaling factor based on the end-to-end delay and reallocates network resources based on the delay scaling factor. If the latency requirements are met, continuous monitoring is performed; In step S1, the arrival process parameters of network slice m include the average arrival rate of data packets of service m and the burst parameters of the arrival process. The system pre-statistically analyzes the arrival of service data packets to obtain the average arrival rate; the service process parameters of network slice m include the computing resource weight factor of the service service process and the burst parameters of its service process. The computing resource weight factor is obtained by calculating the ratio of the computing resources allocated by the current service node to the service to the total computing resources on the service node.

2. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 1 is characterized in that: In step S1, the process of obtaining the arrival process parameters of network slice m is as follows: The service flows are kept independent and identically distributed. The process of service data packets from the service terminal to the base station BS is modeled as a Poisson process. The arrival process of service m arriving during the period [τ, t] is: Among them, a m (i) represents the data packet size of service m from the service terminal to the BS in the i-th time slot, in bits; The Poisson arrival process distribution function is P[N(t)=k]=e -λt (λt) k / k! ,λ m is the average arrival rate of data packets for service m, N(t) represents the event of data packet arrival in time slot t, k represents the number of data packets that arrive in time slot t, and the moment generating function of the Poisson process is: Where θ represents a free parameter, θ>0; Given a constant packet size of v, the cumulative arrival process is A(t) = N(t)v, and we get A m The moment generating function of (τ, t) is: Among them, E[·] represents the expected calculation, A N (τ, t) represents the Poisson arrival process of service m arriving during the period [τ, t]; For random network calculus problems, the affine upper bound of the envelope is achieved: Using affine to reach envelope parameters and Restricted Arrival Process A m (τ,t) When the free parameter δ is greater than 0, the moment generating function and the exponentially bounded burst model are combined with the Chernoff bound inequality to obtain the following equation: Among them, ε A is the probability of violation during the arrival process; Then we get the arrival envelope burst parameters 3. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 2 is characterized in that: In step S1, the upper bound of the delay of network slice m is analyzed and calculated based on the random network calculus theory, including using the arrival envelope parameter and the service envelope parameter, calculating the moment generating function of the service process through the random network calculus model, and using the service cascade theorem and the affine service envelope to calculate the upper bound of the delay of network slice m in different network domains; wherein, The cumulative number of bits processed during the time slot [τ, t] is the service process S m (τ,t), which is expressed as: c m (i) is the time-frequency resource blocks (RBs) and computing resources allocated by the BS to service m in time slot i, that is, the cumulative number of bits that slice m can process; The wireless channel is modeled as a Rayleigh fading channel, and the distance from the BS to the service terminal m is d m The path loss of the transmission channel is Where α is the path loss exponent, and the instantaneous signal-to-noise ratio of service m transmitted in time slot i is: Among them, P m Indicates the user equipment transmit power, P N represents the noise power; g is the channel gain, whose probability density function obeys the exponential distribution. The channel gain probability density function is expressed as: Where λ g represents the expectation of the exponential distribution, and x represents the specific value of the channel gain; The channel gain of each time slot is independent and identically distributed, and the channel transmission service process is expressed as: Where B m Indicates the RB resources allocated to service m; Its moment generating function is: and: Then the channel transmission service process moment generating function is: Among them, define the variable Then we get the channel transmission service envelope parameters Burst Parameters When the service data is transmitted to the BS, the data packet is processed by the MEC server, and the MEC server is modeled as: The total computing power is C MEC , Φ m represents the computing resource weight factor of business m, and σ represents the business surge; The moment generating function of MEC computing service is: From this we get RAN slice domain total service process S m (τ,t) is given by and It consists of two parts, according to the service cascade theorem Its moment generating function is expressed as: Among them, the free parameters θ, δ> 0, the affine service envelope upper bound Setting the affine service envelope parameters and Restriction Service Process S m (τ,t); Under the condition that the wireless transmission service process and the MEC computing service process are independent of each other, if but: like but: Get the affine service envelope parameters K is a temporary variable; ε S represents the probability of service violation, and further obtains the service envelope burst parameter 4. The method for end-to-end network slicing delay balancing based on stochastic network calculus according to claim 3 is characterized in that: In step S2, the burst parameters of the arrival envelope and the service envelope, and the rate of the service envelope are substituted into the data packet delay budget formula to calculate the upper bound of the delay budget for service m: Among them, ε m,RAN represents the delay violation probability of the RAN slice, ε m,RAN =ε A +ε S .

5. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 4 is characterized in that: In step S3, the end-to-end network slice network resources include resource requirements in the RAN domain and the CN domain, specifically including time-frequency resources, computing resources, and link bandwidth resources. Resource allocation is optimized based on service priority while meeting latency requirements. The resource allocation process includes: S30: allocating network resources for end-to-end network slices in the RAN domain and S31: allocating network resources for end-to-end network slices in the CN domain; In step S30, the upper bound of the delay budget W of RAN slice m is m Take the upper bound of the maximum latency of the RAN slice, that is, W m =τ′ m,RAN ,Calculate the minimum amount of resources required for deterministic delay transmission,RAN slice resource allocation includes the following steps: S301: The RAN slice controller obtains the deterministic performance indicators, average packet arrival rate, and MEC server computing capacity of slice m. The deterministic performance indicators include the latency and reliability of slice m. S302: Evenly distribute the available time-frequency resource blocks and computing resources to all RAN slices, and record the resource blocks and computing resources initially allocated to each RAN slice; S303: sorting the slice set in descending order according to the service priority, and allocating resources to services with higher priority first; S304: Determine whether all slice requests have been processed. If so, stop the iteration; otherwise, continue to execute S305; S305: Calculate the number of resource blocks required based on the service transmission requirements and analyze the upper bound of the latency of the RAN slice. S306: The upper limit of the delay is set as the delay requirement, and the required computing resources are determined; S307: Determine whether the calculated required resources exceed the initially allocated resources. If so, repeat S305 and S306 to increase or decrease resource blocks and computing resources to ensure that the service latency requirements are met. S307: Update the remaining time-frequency resource blocks and computing resources in the resource pool, and record the resource allocation plan for slice m.

6. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 5 is characterized in that: In step S31, the following steps are specifically included: The CN slice controller needs to obtain the optimal physical link of the virtual network node VNF deployed in the domain to transmit service m, and then allocate the resources required to ensure deterministic latency based on the selected link. Among them, the KSP algorithm in graph theory is used to sort multiple physical links according to performance indicators, and several better links are selected as candidate paths; For the CN domain, the Internet traffic distribution follows the Gaussian distribution A(t)~(μt,σ 2 t), the latency requirement τ′ of a given CN network slice m m,CN , packet transmission probability p and packet traffic arrival distribution; Let B i is the number of packets that the virtual network node N must process per second or the link E i The number of packets that must be transmitted per second, in bits, where i∈{1,…,k}, k=N+E; let ψ i For virtual network node N or link E i The computing resources or bandwidth resources required to process or transmit a data packet; On the premise of determining the virtual network nodes in the transmission link of slice m, the set of data packets that slice m needs to process or transmit is expressed as: Β={Β1,Β2,…,Β k }, the minimum resource amount is expressed as: The parameters All B i can all be represented by B1, and the optimal value of B1 is found by using the binary search algorithm; Based on the minimum amount of resources, the CN slice controller determines the actual resource requirements initially allocated to each virtual network node on each link.

7. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 6 is characterized in that: In step S4, a cross-domain coordinator is introduced to handle end-to-end slices related to joint resource allocation. The specific process is as follows: When a user or enterprise has a business request, they order a slice from the network service provider (NSP) and submit a related SLA request. After receiving the request, the NSP converts the service demand into slice demand through the cross-domain coordinator. The cross-domain coordinator is responsible for slice orchestration, deployment and maintenance, and decomposes the SLA indicators into two sub-slice SLA indicators for the RAN domain and the CN domain. The sub-domain controller deploys the sub-slices according to the SLA indicators to ensure deterministic latency transmission of the service and continuously monitors the sub-slices, and feeds back the SLA indicators to the cross-domain coordinator in real time.

8. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 7 is characterized in that: In step S4, the end-to-end slice cross-domain coordinator obtains the transmission delay in real time and compares the actual end-to-end delay with the delay requirement, including: S411: The end-to-end slice cross-domain coordinator obtains the transmission delay of the network slice in each domain in real time; the end-to-end slice cross-domain coordinator uses network monitoring tools and probes to collect the transmission delay of the network slice in the RAN slice and CN slice in real time; S412: Compare the actual transmission delay obtained with the preset delay requirement to evaluate the delay performance of the service; obtain the preset transmission requirements of each service from the SLA, including the maximum tolerable delay T m and transmission reliability ε m , the transmission delays in each domain are summarized, and the end-to-end slice cross-domain coordinator calculates the end-to-end transmission delay of each network slice; S413: For services with excessive latency, the end-to-end slice cross-domain coordinator generates warning information, including the ID of the excessive service, the degree of excess, and the time of excess; records the information of the excessive service in the log and archives it for future reference. The recorded content includes: service ID, actual transmission latency, preset latency requirement, and degree of excess; according to the latency excess situation, the end-to-end slice cross-domain coordinator adjusts resource allocation based on the latency ratio factor.

9. The end-to-end network slice delay balancing method based on stochastic network calculus according to claim 8, characterized in that: In step S4, if it is necessary to reallocate network resources according to the latency ratio factor, the process is as follows: S421: Introduce SLA decomposition, and set the latency ratio factor according to the end-to-end latency through the cross-domain coordinator; The cross-domain coordinator defines the delay domain scaling factor χ based on the E2E delay m ∈(0,1), which is used to determine the RAN domain and CN domain delay budgets of slice service m, respectively expressed as: the m,RAN =x m T m the m,CN =(1-x m )T m The total end-to-end latency is: t m,E2E =t m,RAN +t m,CN The E2E scaling factor χ for each service m m Set the initial value and sort the slice set m∈M in descending order of business priority; S422: Dynamically adjust the latency ratio factor according to the latency feedback from each sub-domain controller; After the RAN domain and CN domain controllers obtain the domain ratio factor, use stochastic network calculus and sub-domain slice resource allocation methods to analyze the latency boundary performance and obtain the required amount of allocated resources; Each sub-domain controller performs slice configuration operations according to the allocated resources and records the actual latency; The end-to-end slice cross-domain coordinator calculates the end-to-end transmission delay of each network slice based on the sub-domain controller, and calculates the difference ΔT between the delay requirement and the actual delay m =T m -τ m,E2E ; The cross-domain coordinator determines whether the sum of the latency differences of all slices exceeds the latency difference adjustment threshold. If so, enter step S413; S423: Reallocate network resources to ensure that the actual transmission latency meets the preset latency requirement after the latency ratio factor is adjusted; the cross-domain coordinator determines whether all slice sets have been traversed. If so, record the latency domain ratio factor, otherwise enter step S424; S424: If the actual latency of the RAN slice > RAN latency budget and the actual latency of the CN slice < CN latency budget, increase the ratio factor of the RAN domain and decrease the ratio factor of the CN domain; in, Indicates the upper bound of the delay scale factor; If the actual latency of the RAN slice < RAN latency budget and the actual latency of the CN slice > CN latency budget, decrease the ratio factor of the RAN domain and increase the ratio factor of the CN domain; in, Indicates the lower bound of the delay scale factor; S425: Continuously monitor the adjusted resource allocation plan; The cross-domain coordinator repeats steps S422 - S423 according to the new ratio factor until the sum of the slice latency differences meets the latency difference threshold.

10. A system applicable to the end-to-end network slice delay balancing method based on stochastic network calculus according to any one of claims 1 to 9, characterized in that: It includes: Cross-domain coordinator, RAN slice controller, CN slice controller, monitoring module; among which, The cross-domain coordinator is used to coordinate network resources across the joint domain and perform life cycle management operations for multi-domain slices. The cross-domain coordinator is responsible for converting service requirements into slice requirements, and is responsible for slice orchestration, deployment, and maintenance; RAN slice controller, used to allocate wireless resources and MEC computing resources to perform slice operations on services in the radio access network, ensure service transmission within the RAN domain, monitor resource usage within the RAN domain, and optimize resource allocation by adjusting wireless resource allocation policies; CN slice controller, used to allocate computing resources of service nodes and bandwidth resources on service links, allocate resources according to the best physical link of virtual network nodes for transmitting service m, monitor resource usage within the CN domain, and optimize resource allocation by adjusting the allocation of computing resources and bandwidth resources.

Images