Three-Layer Soft Slicing System and Method Based on Slot Fine Granularity in a Deterministic Network

Abstract

The present invention relates to the technical field of data communication networks. Based on the time-slot fine-grained slicing method, the present invention performs on-demand fine-grained resource slicing of time slots for the deterministic service requirements and the real-time underlying physical network conditions in a deterministic network, ensuring the bandwidth resource allocation and isolation of deterministic services at the fine-grained level, and proposes a three-layer soft slicing system and method based on time-slot fine-grained in a deterministic network. This method can model the deterministic service requirements and network resources within the current time-slot period; calculate the priority of deterministic services by combining QoS metrics such as bandwidth, delay, jitter, and packet loss of deterministic service requirements, then slice and allocate time slots for deterministic services according to the remaining resources of the underlying network, and perform dynamic adjustment according to network feedback. The present invention can maximize the utilization rate of the three-layer resources of the deterministic network while ensuring the service quality of deterministic services as much as possible.

Description

Technical Field

[0001] The present invention relates to the technical field of data communication networks, and more specifically, particularly to a three-layer soft slicing system and method for a deterministic network. Background Art

[0002] In recent years, the Internet has entered the stage of industrial Internet. With the wide attention paid to remote control latency-sensitive services such as telemedicine, remote driving, and network manufacturing, the common requirements of such services for the underlying network include metrics such as bounded bandwidth, latency, jitter, and packet loss. The existing resource reservation mechanism based on service QoS categories is limited by the "statistical multiplexing and best effort" transmission characteristics of the IP datagram network, and can only partially guarantee the service bandwidth, and it is difficult to meet the upper and lower bound guarantee requirements such as latency jitter. In view of this, the industry has proposed a deterministic network for guaranteeing the quality of service (QoS) of latency-sensitive services in large-scale backbone networks, and resource reservation and allocation technologies are used to reduce resource competition generated during the transmission of new services at the network layer.

[0003] Currently, both the industrial community and the academic community have studied resource reservation and allocation technologies in the three layers of the deterministic network. For complex services, the network slicing technology has been proposed to divide multiple virtual end-to-end subnets at the network layer. Today's network slicing can rationally allocate device resources through network function virtualization (NFV), convert them into multiple end-to-end network slices with different granularities and high isolation degrees, and virtualize multiple subnet slices that are respectively adapted to different services and insulated from each other by means of networking on demand. And exclusive resources are allocated to each network slice, so that they can guarantee services with different requirements respectively, the resources between slices are isolated from each other, and the state within one slice will not affect the data transmission of other slices. However, the existing resource slicing schemes in the network are basically to slice each type of service, so as to simply meet the deterministic QoS requirements of each type of service in the deterministic network and improve and optimize the utilization rate of the link. However, this slicing method belongs to coarse-grained slicing. Many only consider slicing through simple classification of services, so as to achieve simple virtualized slicing and isolation between services. Many do not consider the service differences of slices and the fairness between different services. And the resources allocated within the three-layer slice sometimes still have resource competition with the bursts of different deterministic services, resulting in general service quality of deterministic services and resource utilization rate of the link after slicing, which does not conform to the idea of deterministic guarantee of deterministic traffic. To sum up, the current deterministic three-layer soft slicing method has the following problems:

[0004] 1. The fine-grained requirements of each deterministic service are not considered, and only the classification of deterministic services is used to slice and isolate the three layers of the network.

[0005] 2. The burst situation of deterministic traffic and the type of traffic within each type of slice are not considered, and there is still resource competition within the slice, lacking flexibility. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a three - layer soft slicing system and method based on time - slot fine - granularity in a deterministic network. Slicing is performed for each type of deterministic service from the time - slot fine - granularity at the network layer to reduce the resource competition problem when deterministic services arrive, and while ensuring the QoS requirements of each type of deterministic service, the network resource utilization rate is improved as much as possible.

[0007] To solve the above - mentioned technical problem, the present invention proposes a three - layer soft slicing system based on time - slot fine - granularity in a deterministic network, including an information collection module, a network transmission module connected to the input end of the information collection module, and a time - slot control module connected to the output end of the information collection module. The input end of the time - slot control module is the information collection module, and the output end is the network transmission module. The input end of the network transmission module is the time - slot control module, and the output ends are the information collection module and the underlying network topology.

[0008] The information collection module includes a basic requirement collection module for deterministic services and an actual transmission information module for deterministic services; the time - slot control module includes a priority construction module for deterministic services, a fine - granularity time - slot slicing construction module for deterministic networks, and a fine - granularity time - slot slicing scheduling module for deterministic networks; the network transmission module includes a flow table distribution module for deterministic networks and a time - slot slicing policy configuration module.

[0009] Based on the above system, the present invention also proposes a three - layer soft slicing method based on time - slot fine - granularity, including the following steps:

[0010] S1. Use the deterministic network controller in the information collection module to obtain the topology information of the underlying physical network, including network topology routing information, bandwidth and other information, and deterministic service demand information;

[0011] S2. Model based on the traffic demand data of historical deterministic services, where the demand data includes bandwidth, jitter, delay, and packet loss, and input it into the time - slot control module for priority construction before the time - slot slicing module in the priority construction module for deterministic services;

[0012] S3. Considering the network basic information data comprehensively, according to the types and requirements of deterministic services in each period, quantify the three-layer resources into time slots, perform three-layer soft slicing of each type of deterministic service based on time slots with fine granularity in the time slot control module, output the deterministic service fine-grained time slot resource allocation strategy in the three layers of the deterministic network, and send the strategy of performing time slot fine-grained slicing to the network transmission module for the distribution of time slot slicing resource allocation strategy and flow table distribution. In the network transmission module, control the transmission time of each type of deterministic service in the network layer through the time slot slicing and flow table of the deterministic service;

[0013] S4. According to each type of deterministic QoS metric measured by the controller in the information collection module in the deterministic network and the situation of the links in the network layer, dynamically adjust the time slot slicing allocation for each type of deterministic service, recycle or release the corresponding time slot slicing at an appropriate time, and make adjustments when the time slot slicing cannot meet the deterministic requirements, and then return to S2.

[0014] Furthermore, the step S1 specifically includes: obtaining the network basic information and historical deterministic service information from the information collection module, where the network layer basic information includes routing, link, bandwidth, and port information, and the historical deterministic service information includes metrics such as bandwidth, delay, jitter, and packet loss. Since the deterministic service traffic has characteristics such as periodicity and the determination of the upper and lower bounds of basic QoS metrics, the historical deterministic service requirements can represent the basic requirements of the deterministic service.

[0015] Preferably, step S2 includes the following sub-steps:

[0016] S21. Input the deterministic service requirements from the information collection module into the deterministic service priority construction module in the time slot control module. The requirement data includes bandwidth, jitter, delay, and packet loss. After obtaining the information of the deterministic service, model the period and requirement information of each type of deterministic service, and based on the modeled deterministic service, determine the time slot slicing period and time slot granularity, and calculate all the requirements of the deterministic service with all time slot granularities within the period and input them into the time slot control module.

[0017] S22. Perform deterministic service priority construction on all the deterministic services input into the time slot control module within the period T, and input all the deterministic services with priority construction into the time slot slicing module.

[0018] Preferably, step S3 includes the following sub-steps:

[0019] S31. Quantify the time slots for the three-layer network layer bandwidth resources based on the service requests and demand modeling information of each type of deterministic service with priorities.

[0020] S32. Use the greedy algorithm to select deterministic services with priority sorting for time slot slicing. Allocate time slots for this deterministic service during the time slot period T while ensuring deterministic requirements, and allocate all deterministic services to different time slots for transmission, so that their transmission times in the network are separated, achieving the purpose of resource slicing in the three layers of the deterministic network.

[0021] After all deterministic services are allocated, extract the path in the service modeling as the three-layer flow table policy, and extract the time in the service as the three-layer soft slicing policy based on time slot fine-grainedness. Send the two policies to the network transmission module.

[0022] Preferably, step S4 includes the following sub-steps:

[0023] S41. For the time slot slicing situation in the existing network and the deterministic service situation in the network, determine whether the time slot slicing allocation policy can meet the quality of service requirements of the deterministic network.

[0024] S42. If the existing time slot slicing cannot meet the deterministic service requirements and does not meet the constraints such as delay, bandwidth, jitter, and packet loss in the deterministic network, then according to the resource situation in the current network, perform new time slot slicing and resource allocation for this type of deterministic service, and repeat steps S2 - S4. If the time slot can meet the requirements of the deterministic service, then continue to use this fine-grained time slot policy for deterministic service transmission.

[0025] Based on the time slot fine-grained slicing method, the present invention performs on-demand fine-grained resource slicing of time slots for the deterministic service requirements and the real-time underlying physical network situation in the deterministic network, and ensures the bandwidth resource allocation and isolation of deterministic services from the fine-grained level, and proposes a three-layer soft slicing system and method based on time slot fine-grainedness in the deterministic network. This method can model the deterministic service requirements and network resources within the current time slot period; calculate the priorities of deterministic services in combination with QoS indicators such as bandwidth, delay, jitter, and packet loss of deterministic service requirements, then slice and allocate time slots for deterministic services according to the remaining resources of the underlying network, and perform dynamic adjustment according to network feedback. The present invention can maximize the resource utilization rate of the three layers of the deterministic network while ensuring the quality of service of deterministic services as much as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The technical solutions of the present invention will be further specifically described below in conjunction with the drawings and specific embodiments.

[0027] Figure 1 It is a framework diagram of a three-layer soft slicing system based on time slot fine-grainedness of the present invention;

[0028] Figure 2A diagram showing the relationship between the modules of the three-layer soft slice system based on time slot fine granularity of the present invention;

[0029] Figure 3 It is an overall flow chart of the time slot fine-grained slicing method of the present invention;

[0030] Figure 4 This is a flow chart of time slot fine-grained slicing based on genetic algorithm of the present invention.

[0031] Figure 5 This is a flowchart of the steps of the time slot-based fine-grained slicing method of the present invention. DETAILED DESCRIPTION

[0032] Combination Figure 1 , 2 As shown, a three-layer soft slicing system based on time slot fine-grainedness in a deterministic network includes the following modules: time slot control module, information collection module, and network transmission module.

[0033] The information collection module includes a deterministic business basic demand collection module and a deterministic business actual transmission information module; the time slot control module includes a deterministic business priority construction module, a deterministic network fine-grained time slot slice construction module, and a deterministic network fine-grained time slot slice scheduling module; the network transmission module includes a deterministic network flow table delivery module and a time slot slice strategy configuration module.

[0034] The input end of the information acquisition module is the network transmission module, and the output end is the time slot control module. The information acquisition module is mainly used to measure basic information in the deterministic network. The specific packet sending and measurement use the Distributed Internet Traffic Generator (D-ITG), which supports the generation of Internet Protocol Version 4 and Internet Protocol Version 6 traffic, and can generate multi-layer traffic such as network layer, transport layer and application layer. It has a variety of parameters, which can be customized and adjusted as needed, and can measure QoS data such as bandwidth rate, delay, jitter, packet loss rate, etc. of the service. The measurement content includes the following two: one is the basic information in the network, including router connection status, port status, link initial bandwidth status, etc.; the other is to measure real-time information in the network, including the current link load status and network traffic status, and obtain the bandwidth, delay, and jitter information required for deterministic traffic and store them in the database.

[0035] The input end of the time slot control module is the information collection module, and the output end is the network transmission module. The time slot control module mainly collects the basic demand information of deterministic services from the information collection module, including the requirements for deterministic service bandwidth, delay, jitter, packet loss, etc. The deterministic service priority construction module determines the cycle granularity and cycle, quantifies the network bandwidth within the cycle into time slots, and constructs the priorities for the requirements of deterministic services; then, the genetic algorithm is used for time slot slicing and resource reservation in the current cycle, and the start transmission time and transmission path of all deterministic network services in the current cycle are output. The time slot scheduling module measures the required bandwidth, delay, and jitter of the real-time deterministic traffic in the network output by the measurement module. If the requirements of the deterministic service are not met, the time slots are modified to perform resource scheduling of the time slots in the deterministic network slice, and a slice reconstruction based on fine-grained time slots is performed.

[0036] The input end of the network transmission module is the time slot control module, and the output ends are the information collection module and the underlying network topology, including deterministic switches, deterministic virtual switching software, and network topology. The modules included in the network transmission module are: the deterministic network flow table distribution module, which configures the ports of the network layer and distributes the flow tables through the flow table information. The deterministic network time slot configuration module allocates time slots for each type of service through the time slot information of each type of deterministic service output by the time slot control module. Specifically: if the current service has a transmission time in this time slot, it is transmitted according to the flow table information; if the deterministic service has no transmission time in the current time slot, it waits for the next time slot.

[0037] As Figure 5 shown, the three-layer soft slicing method based on fine-grained time slots includes the following steps:

[0038] S1. Use the deterministic network controller in the information collection module to obtain the topology information in the underlying physical network, including network topology routing information, bandwidth, etc. information and deterministic service demand information;

[0039] S2. Based on the traffic demand data of historical deterministic services, perform modeling. The demand data includes bandwidth, jitter, delay, and packet loss, and input it into the time slot control module for priority construction before the time slot slicing module in the deterministic service priority construction module;

[0040] S3. Considering the network basic information data comprehensively, according to the types and requirements of deterministic services in each period, quantify the three-layer resources into time slots, perform three-layer soft slicing of each type of deterministic service based on time slot fine-grained in the time slot control module, output the fine-grained time slot resource allocation strategy of deterministic services in the three layers of the deterministic network, and send the strategy of performing time slot fine-grained slicing to the network transmission module to perform the distribution of time slot slicing resource allocation strategy and flow table distribution. In the network transmission module, control the transmission time of each type of deterministic service in the network layer through the time slot slicing and flow table of the deterministic service;

[0041] S4. According to each type of deterministic QoS metric measured by the controller in the information collection module in the deterministic network and the situation of the links in the network layer, dynamically adjust the time slot slicing allocation for each type of deterministic service, recycle or release the corresponding time slot slicing at the appropriate time, and make adjustments when the time slot slicing cannot meet the deterministic requirements, and return to S2.

[0042] Step S1 specifically includes: using a deterministic controller in the information collection module to collect the basic information and historical deterministic service demand information in the three layers of the deterministic network, where the network layer basic information includes routing, link, bandwidth, and port information, and the historical deterministic service information includes metrics such as bandwidth, delay, jitter, and packet loss. Since the deterministic service traffic has characteristics such as periodicity and the determination of the upper and lower bounds of basic QoS metrics, the historical deterministic service demand can represent the basic demand of the deterministic service.

[0043] Combined Figure 1 、 2 As shown, step S2 includes the following sub-steps:

[0044] S21. Input the deterministic service demand from the information collection module into the deterministic service priority construction module in the time slot control module. The demand data includes bandwidth, jitter, delay, and packet loss. After obtaining the information of the deterministic service, model the period and demand information of each type of deterministic service, and based on the modeled deterministic service, determine the time slot slicing period and time slot granularity, and calculate all the demands of the deterministic service with all time slot granularities within the period and input them into the time slot control module.

[0045] S22. Perform deterministic service priority construction on all the deterministic services input into the time slot control module within the period T, and input all the deterministic services with priority constructed into the time slot slicing module.

[0046] Step S21 specifically includes modeling all the deterministic services, modeling all the deterministic service flows within the time slot period as a set F, where the flow f in F i The modeling is as follows:

[0047] f i=(src, dst, cycle, type, bandwidth, start, time, path)

[0048] represents a certain deterministic traffic flow f within the period T i The sending node is src, the destination node is dst, cycle represents the traffic flow period, the bandwidth requirement of this traffic flow is bandwidth, start is the time when the traffic flow arrives at the network, time is the start transmission time, and path is the transmission path. Among them, the first five are known quantities, and time and path are unknown quantities, which are the flow table strategy and time slot slicing strategy output by the time slot construction module.

[0049] For the i-th deterministic service, its relative deterministic requirement Q i is modeled as follows:

[0050] Q i =(d_max, j_max, l_max, b_min, b_max)

[0051] represents that the upper bound of the delay of the i-th deterministic service is Q i .d_max, the upper bound of jitter Q i .j_max, the upper bound of packet loss Q i .l_max, the lower bound of bandwidth Q i .b_min and the upper bound of bandwidth Q i .b_max.

[0052] After modeling the service requests and demands of all deterministic services, calculate the least common multiple T of the periods of all deterministic services according to the service request F, calculate the greatest common divisor N of the periods of all deterministic services, and use T as the time slot period and N as the time slot granularity.

[0053] Step S22 specifically includes constructing priorities for all deterministic services. When performing time slot slicing later, ensure that time slot slicing is performed within the range of the upper bound of the delay of the deterministic service. For the priorities of deterministic services, according to the upper bound of the delay Q i .d_max in the QoS requirement minus the arrival time f i .start is used as the priority of the time slot allocation algorithm. Then, allocate the service with the highest priority among the currently unprocessed deterministic services in the current time slot. For deterministic services with relatively low priorities, if they cannot be allocated in the current time slot, since all services are shifted to the next time slot, directly allocate this service in the next time slot. The grade is calculated as follows.

[0054] grade = Q i .d_max - f i.start

[0055] After obtaining all the deterministic service priorities, the input time slot construction module performs time slot slicing.

[0056] Combined as Figure 3 , 4 shown, step S3 includes the following sub-steps:

[0057] S31. Based on the service requests and demand modeling information of each type of deterministic service with priorities, quantify the time slots for the three-layer network layer bandwidth resources.

[0058] S32. Use the greedy algorithm to select the deterministic services with priority sorting for time slot slicing, allocate time slots for this deterministic service within the time slot period T while ensuring deterministic requirements, allocate all deterministic services to different time slots for transmission, so that their transmission times in the network are separated, achieving the purpose of resource slicing in the three layers of the deterministic network.

[0059] After the allocation of all deterministic services is completed, extract the path in the service modeling as the three-layer flow table policy, and extract the time in the service as the three-layer soft slicing policy based on time slot fine granularity. Send the two policies to the network transmission module.

[0060] Step S31 specifically includes that after having the time slot period T and time slot granularity N, quantify the time slots for the three-layer resources. Among them, the resources in the three layers are currently mainly bandwidth, so quantify the bandwidth. For the three-layer network, it can be described as an undirected graph G=(V, E), where V represents the set of physical nodes V=(v1, v2, v3, …, v d ), d represents the number of nodes; E represents the set of physical links in the network E=(e1, e2, e3, …, e r ), r represents the number of links. Physical nodes are connected by links. Each physical link e s , 1≤s≤d can also be written as e (i,j) , composed of vertices i, j, and the link bandwidth resource is B(e s ).

[0061] Therefore, to quantify the link bandwidth resources, the bandwidth granularity bw(a) corresponding to each time slot granularity within the period T is as follows.

[0062] bw(a) = BW(a) * T / N

[0063] According to the conventional expression in this field, BW(a) is the bandwidth granularity corresponding to all time slot granularities within one period of relative bw(a). If the bandwidth requirement in the link period T of this deterministic service request is b(a), then the time slot resources n allocated within the period T in the deterministic network are as follows.

[0064] n = b(a) / bw(a)

[0065] Step S32 specifically includes using the genetic algorithm to perform time slot allocation and time slot reservation for each deterministic service within the time slot period. During time slot assembly, for each deterministic service, its deterministic requirements need to be met during the assembly process. For a deterministic service, if it is allocated to the next time slot, queuing time will be involved, and the waiting time w_time should be added in terms of delay. Suppose there are m deterministic services within the T period, and the set of all deterministic services from the source node src to the destination node dst is F m , there are n m service flows in total, and the i-th service from the source node src to the destination node dst in F m is f i (src, dst), w_time is the sending time of service flow i minus the arrival time, represents the delay of the service f i (src, dst) in the network from the source node src to the destination node dst, represents the jitter of the service f i (src, dst) in the network from the source node src to the destination node dst, represents the packet loss rate of the service f i (src, dst) in the network from the source node src to the destination node dst, represents the bandwidth requirement of the service f i (src, dst) in the network from the source node src to the destination node dst. Suppose the upper bound of the delay of the m-th type of deterministic service is Q m .delay, the upper bound of the jitter is Q m .jittle, the upper bound of the packet loss rate is Q m .loss, and the lower bound of the bandwidth is Q m .b_min. The constraint conditions of the m-th type of deterministic service are as follows:

[0066]

[0067] For deterministic services, the earlier they are assembled, the less the queuing time, and the better the optimization of latency. Using as few resources as possible during the assembly process for allocation can enable more service allocation for the time slot resources. Therefore, the goal of the genetic algorithm for each deterministic service time slot slice is to minimize the three-layer time slot resource usage. Assume service f i (src, dst) uses the resource amount during assembly in the entire deterministic network slice as Then the optimization goal is as follows.

[0068]

[0069] After having the optimization conditions and the objective function, the genetic algorithm is used to perform time slot slicing for each deterministic service. The main process is to input the deterministic services to be allocated, establish a time slot allocation population, and then perform fitness evaluation on the population. If the specified number of iterations of the genetic algorithm is not reached, the next generation of the population is generated through individual selection, gene crossover, gene mutation, etc. In this process, individuals with low fitness are eliminated, and finally the optimal time slot allocation solution is output. Therefore, the time slot scheduling method based on the genetic algorithm mainly includes the encoding and decoding, fitness function, termination condition, and individual selection operation, gene crossover operation, gene mutation operation in the initialization operation, which are defined as follows:

[0070] 1) Encoding and decoding

[0071] Before performing the genetic algorithm, the parameters in the network need to be encoded. There are two ways of network encoding in the genetic algorithm: binary encoding of edges and priority encoding of nodes. Since each service request is an end-to-end transmission from the source node to the destination node, encoding the edges will generate a large number of infeasible solutions and consume too much time and resources. Therefore, the encoding scheme adopted is to encode the nodes with priorities, and the priority sequence numbers of each node are randomly generated during encoding.

[0072] For a deterministic service flow to be allocated, it is necessary to calculate the time time when it is deployed and the path path of the deployed flow. Therefore, sh i ={time[i], path[i]} is used to represent any chromosome of the encoding, indicating any deterministic service flow f i , time[i] represents the time slot to be allocated, and path[i] represents the path information of the service flow f i that is allocated. Among them, path[i] consists of paths, path[i] = {src, next1, next2..., dst}, and a link in the network can be formed by two adjacent nodes in path[i].

[0073] When generating the initial population and individual encoding, first, it is necessary to determine whether to allocate resources in this time slot according to the priority decision of this service and the remaining resources of the time slot. If it is selected to allocate in this time slot, it is necessary to generate a path for it. In order to make the population generated each time have strong randomness, a random routing algorithm is used to generate the path. Then, based on this information, it is compared with the existing time slot allocation strategy in the network. When there is no conflict and the deterministic service delay requirement is met, the initial population is generated.

[0074] The random routing algorithm randomly generates the serial numbers of each node before the algorithm starts when generating the path. When selecting the path, it searches for the node with the largest serial number and connected links as the next node. Each time the algorithm performs time slot allocation and genetic iteration for the deterministic service, it calls the random routing selection algorithm to ensure that different paths are generated for the same deterministic service in the network, so as to meet the requirement of population diversity.

[0075] In the decoding process, it is necessary to allocate traffic on the paths from randomly generated different source nodes to the destination node, and the laws such as traffic conservation need to be satisfied during this process. The paths that cannot be allocated traffic are deleted, and finally the links that meet the conditions in this time slot are found. The specific process is as follows: First, generate the path Path[i] through the random priority generation and random routing algorithm. Then, calculate the minimum traffic min_path among all paths in Path[i], and set the path traffic to the minimum traffic min_path. Judge whether the minimum traffic meets the traffic demand of the deterministic service flow. If it is less than its service demand, regenerate the random path for another iteration. If it is greater than its service demand, it means that it can be allocated to this link, then add it to this link, and update the deterministic service path. Then update the capacity of each link in Path[i] in the whole network, and return multiple paths as the initial population finally.

[0076] 2) Fitness function

[0077] In the genetic algorithm, the fitness function is used to evaluate the fitness of each individual in the current population. The value of the fitness can be used as the basis for whether this individual will be selected. If an individual has a high fitness, it has a greater probability of entering the next generation population. Therefore, for each chromosome, the previously established optimization objective function can be directly used as the fitness function of the time slot allocation method based on the genetic algorithm.

[0078] 3) Termination condition

[0079] In the genetic algorithm, the termination conditions of the algorithm mainly include reaching the objective function, reaching the number of iterations, and the iterative solution being unable to be further optimized. Since the fitness function is a minimization optimization problem and the specific objective function cannot be reached, the condition that the number of iterations reaches the set value is used to judge the termination of the algorithm.

[0080] 4) Individual selection

[0081] For the individual selection operation, since the fitness function is a minimization problem and the number of deterministic services to be allocated is large, compared with roulette wheel selection, tournament selection with faster convergence speed and stronger versatility is usually used. The main idea is as follows: Repeat k times to simulate the tournament method, randomly select tournise individuals from the population with equal probability, calculate their fitness, and select the individual with the best fitness to enter the next population.

[0082] 5) Gene crossover

[0083] Gene crossover in the genetic algorithm is a process that starts from the genes of the parent generation, performs operations, and finally obtains the genes of the offspring generation. Since the initially generated path is a subsequence, the ordered crossover strategy is adopted for the crossover operation. The specific steps of the ordered crossover operation are as follows:

[0084] (1) Find two parent genes, find two different paths Path[i] of the deterministic service f_i from the parent generation, then compare the minimum value min of the two parent paths, and then randomly generate two values a and b (a < b) starting from 2 to min as the starting position a and ending position b of the random start of the parent sequence;

[0085] (2) Use the sequence starting from position a and ending at position b of the parent generation as the subsequence. Then find the order of the remaining sequence from the other parent and insert it into the subsequence;

[0086] (3) Finally, form the sequence genes of the two offspring generations.

[0087] 6) Gene mutation

[0088] Gene mutation in the genetic algorithm is also a process that starts from the genes of the parent generation, performs operations, and finally obtains the genes of the offspring generation. Among them, gene mutation and gene crossover can be called gene mutation algorithms. The gene mutation of this algorithm adopts the disordered mutation strategy, where disordered mutation randomly shuffles the values in the sequence, and the probability value mutpb gives whether the coding position in each gene changes.

[0089] 7) Environmental selection

[0090] Environmental selection means that after selection, crossover, and mutation, the population size obtained may increase or decrease compared to the parental generation. To maintain the population size, it is necessary to insert the breeding offspring into the parental generation, replace a part of the individuals in the parental population, or discard a part of the breeding individuals. Since binary tournament selection is used for individual selection in the algorithm and the number of cycles is equal to the population size, the population size after crossover and mutation will not change compared to the parental generation. Therefore, the environmental selection of the algorithm is complete reinsertion without elitist retention. Finally, the paths of each deterministic service and the allocated time slot data are calculated through the genetic algorithm.

[0091] Step S33 specifically includes extracting each deterministic service path path and allocated time slot time generated by the genetic algorithm based on time slot fine-grainedness, integrating all paths within the time slot period T to form a flow table policy. Integrating the allocated time slot data of all services to form a three-layer soft slicing policy based on time slot fine-grainedness, and sending the two policies to the network transmission module for control and network transmission. Resource scheduling for deterministic services is carried out through the transmission policy sent by the control layer, and the transmission time of deterministic services is controlled to ensure that the transmission times of deterministic services in the three layers of the deterministic network are different. Achieving time slot fine-grained slicing and resource isolation in the three layers of the deterministic network.

[0092] Combined Figure 3 、 Figure 4 As shown, step S4 includes the following sub-steps:

[0093] S41. For the time slot slicing situation and the deterministic service situation in the existing network, judge whether the time slot slicing allocation policy can meet the quality of service requirements of the deterministic network.

[0094] S42. If the existing time slot slices cannot meet the deterministic service requirements and do not meet the constraints such as delay, bandwidth, jitter, and packet loss in the deterministic network, then according to the resource situation in the current network, new time slot slicing and resource allocation are carried out for this type of deterministic service, and steps S2 - S4 are repeated. If the time slot can meet the deterministic service requirements, then continue to use this fine-grained time slot policy for deterministic service transmission.

[0095] Step S41 specifically includes: Through the information collection module, QoS feedback such as delay, jitter, packet loss, and bandwidth for deterministic services is carried out. It is analyzed that the bandwidth of the deterministic traffic is WL%, the delay is DL ms, the jitter is JL ms, and the packet loss is LL%. Then calculate whether the indicators of the deterministic traffic meet the constraints of the deterministic service.

[0096] Step S42 specifically includes: After obtaining that the bandwidth of each type of deterministic traffic is WL%, the delay is DL ms, the jitter is JL ms, and the packet loss is LL% through S41, if the QoS requirements of this type of deterministic traffic are not met, it means that this time slot slice needs to be adjusted. Then, the construction and generation of the time slot fine-grained slice are performed again for the non-compliant deterministic services. Repeat the steps of S2 - S4 until all deterministic services meet the QoS requirements.

[0097] Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention rather than 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 spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A three-layer soft slicing system based on time-slot fine-grained in a deterministic network, characterized in that It includes an information acquisition module, a network transmission module connected to the input end of the information acquisition module, and a time slot control module connected to the output end of the information acquisition module; the input end of the time slot control module is the information acquisition module, and the output end is the network transmission module; the input end of the network transmission module is the time slot control module, and the output ends are the information acquisition module and the underlying network topology; The information acquisition module is mainly used for measuring the basic information in the deterministic network. The information acquisition module includes a deterministic service basic requirement acquisition module and a deterministic service actual transmission information module; the measurement content includes the following two: one is the basic information in the network, including router connection status, port status, and initial link bandwidth status; the other is to measure the real-time information in the network, including the current load status of the link and the network traffic status, and store the bandwidth, delay, and jitter information required for deterministic traffic in a database; The time slot control module includes a deterministic service priority construction module, a deterministic network fine-grained time slot slicing construction module, and a deterministic network fine-grained time slot scheduling module; the deterministic service priority construction module determines the cycle granularity and cycle, quantifies the network bandwidth within the cycle into time slots, and constructs the priority for the requirements of deterministic services; the deterministic network fine-grained time slot slicing construction module performs time slot slicing and resource reservation on the current cycle using a genetic algorithm, and outputs the start transmission time and transmission path of all deterministic network services in the current cycle; the deterministic network fine-grained time slot scheduling module measures the required bandwidth, delay, and jitter of the real-time deterministic traffic in the network output by the measurement module. If the requirements of the deterministic service are not met, it modifies the time slots to perform resource scheduling of the time slots in the deterministic network slice, and performs re-construction of the time slot based on fine-grained slicing; The network transmission module includes a deterministic network flow table distribution module and a time slot slicing strategy configuration module; the deterministic network flow table distribution module configures the ports and distributes the flow tables to the network layer through the flow table information; the time slot slicing strategy configuration module performs time slot allocation for each type of service through the time slot information of each type of deterministic service output by the time slot control module.

2. A three-layer soft slicing method using the three-layer soft slicing system described in claim 1, characterized in that, It includes the following steps: S1. Use the deterministic network controller in the information acquisition module to obtain the topology information of the underlying physical network, including network topology routing information, bandwidth information, and deterministic service requirement information. The deterministic service requirement information includes deterministic service bandwidth, delay, jitter, and packet loss; S2. Perform modeling based on the traffic demand data of historical deterministic services. The demand data includes bandwidth, jitter, delay, and packet loss, and input it into the time slot control module for priority construction before the time slot slicing module in the deterministic service priority construction module; S3. Considering the network basic information data comprehensively, according to the types and requirements of deterministic services in each period, quantify the three-layer resources into time slots, perform three-layer soft slicing of each type of deterministic service based on time slot fine-grained in the time slot control module, output the deterministic service fine-grained time slot resource allocation strategy in the three layers of the deterministic network, and send the strategy of performing time slot fine-grained slicing to the network transmission module to perform the distribution of the time slot slicing resource allocation strategy and the flow table distribution. In the network transmission module, control the transmission time of each type of deterministic service in the network layer through the time slot slicing and flow table of the deterministic service; S4. According to the deterministic QoS metrics of each type of service measured by the controller in the information collection module in the deterministic network and the situation of the links in the network layer, dynamically adjust the time slot slicing allocation for each type of deterministic service, recycle or release the corresponding time slot slices at an appropriate time, and make adjustments when the time slot slices cannot meet the deterministic requirements, and return to S2.

3. The three-layer soft slicing method according to claim 2, wherein, The specific steps of step S1 include: obtaining the network layer basic information and historical deterministic service information from the information collection module, where the network layer basic information includes routing, links, bandwidth, and port information, and the historical deterministic service information includes bandwidth, delay, jitter, and packet loss metrics.

4. The three-layer soft slicing method according to claim 2, characterized in that, Step S2 includes the following sub-steps: S21. Input the deterministic service requirements from the information collection module into the deterministic service priority construction module in the time slot control module. The requirement data includes bandwidth, jitter, delay, and packet loss. After obtaining the information of the deterministic service, model the period and requirement information of each type of deterministic service, and determine the time slot slicing period and time slot granularity based on the modeled deterministic service. Calculate all the requirements of the deterministic service for all time slot granularities within the period and input them into the time slot control module; S22. Construct the deterministic service priorities for all the deterministic services input into the time slot control module within the period T, and input all the deterministic services with the constructed priorities into the time slot slicing module.

5. The three-layer soft slicing method according to claim 2, wherein Step S3 includes the following sub-steps: S31. Quantify the time slots for the three-layer network layer bandwidth resources based on the service requests and demand modeling information of each type of deterministic service with priorities; S32. Use the greedy algorithm to select the deterministic services with priority sorting for time slot slicing, allocate time slots for the deterministic service within the time slot period T while ensuring the deterministic requirements, and allocate all the deterministic services to different time slots for transmission, so that their transmission times in the network are separated, achieving the purpose of resource slicing in the three layers of the deterministic network; S33. After the allocation of all the deterministic services is completed, extract the path in the service modeling as the three-layer flow table strategy, and extract the time in the service as the three-layer soft slicing strategy based on time slot fine-grained; send the two strategies to the network transmission module.

6. The three-layer soft slicing method according to claim 2, wherein Step S4 includes the following sub-steps: S41. For the existing time slot slicing situation in the network and the situation of the deterministic services in the network, judge whether the time slot slicing allocation strategy can meet the service quality requirements of the deterministic network; S42. If the existing time slot slices cannot meet the requirements of deterministic services and do not meet the deterministic network delay, bandwidth, jitter, and packet loss constraint conditions, then according to the resource situation in the current network, new time slot slices and resource allocation are performed for this type of deterministic service, and steps S2 - S4 are repeated; if this time slot can meet the requirements of deterministic services, then continue to use this fine-grained time slot strategy for deterministic service transmission.

Images