A VNF deployment method based on coupled bandwidth allocation and with delay QoS guarantee

Abstract

The present invention belongs to the technical field of network function virtualization, and specifically relates to a VNF deployment method based on coupled bandwidth allocation and with delay QoS guarantee; abstracting a physical network, obtaining a SFC to be deployed, and pre-allocating resources according to the characteristics of arriving traffic based on queuing theory and large deviation theory; sorting physical nodes in descending order according to the size of remaining CPU resources, screening candidate physical nodes and candidate physical links respectively according to pre-allocated CPU resources and bandwidth resources, estimating the coupled bandwidth after coupling new traffic with existing traffic of the candidate link based on martingale theory, and solving the coupled bandwidth increment of the candidate link; solving the joint optimization problem of satisfying delay QoS and minimum resource occupancy rate, and obtaining the optimal deployment plan; the method can allocate resources according to the arrival characteristics of data packets requested by the service, and achieve delay QoS guarantee while minimizing resource occupancy rate; in addition, the method can also fully save network resources and achieve the same SFC carrying capacity with less consumption.

Description

Technical Field

[0001] The present invention belongs to the technical field of network function virtualization, and in particular relates to a VNF deployment method based on coupled bandwidth allocation and with delay QoS guarantee. Background Art

[0002] With the gradual acceleration of 5G commercialization in recent years, the three major application areas of "low-latency high-reliability communication", "large-scale Internet of Things" and "mobile enhanced bandwidth" have gradually enriched, resulting in a sharp increase in the types and number of services. In order to provide differentiated services for diverse services, service providers need to flexibly configure network functions and network resources to meet the QoS (quality of service) requirements of the services. The application of NFV (network function virtualization) and SDN (software defined network) technologies has enabled 5G networks to achieve software and hardware decoupling, control plane and data plane decoupling, significantly enhancing the flexibility of 5G networks and making 5G networks customized and programmable. With the help of NFV and SDN, 5G slicing technology cuts and reorganizes the limited physical resources in the network to form a resource pool for on-demand calls by services within the slice, forming a logically isolated slice network, reducing the cost of network operation while achieving efficient resource utilization.

[0003] In the 5G slicing network, each service request is composed of multiple VNFs (virtual network functions) to form a virtual link, called SFC (service function chaining). Services with different requirements form different SFCs. The network controller allocates resources according to the SFC of the service requests arriving in the network, realizes the mapping of virtual links to physical links, plans a suitable physical path for the data packets arriving in the service, and divides the required network resources, such as CPU resources or bandwidth resources. This process is called VNF deployment. How to maximize benefits or minimize consumption while meeting service needs through VNF deployment has always been a key issue in 5G research.

[0004] Through VNF deployment, the network realizes flexible allocation and fine-grained control of resources. At the same time, 5G networks have high requirements for latency guarantee. Existing inventions all achieve latency guarantee at the level of physical facilities, modeling single-hop latency as a function related to resource occupancy, and then superimposing multi-hop latency. The prior art focuses on the relationship between single-hop latency and network service process, but the data arrival characteristics of the service are a crucial factor in the transmission process. Therefore, in VNF deployment, how to achieve latency QoS guarantee for services with different arrival characteristics is a very important issue. In addition, the resources within the slice are occupied by multiple services, and how to deal with the resource occupancy relationship between multiple services is also a problem worthy of attention. Summary of the invention

[0005] In order to overcome the above problems, the present invention provides a VNF deployment method based on coupled bandwidth allocation and with delay QoS guarantee, which can allocate resources according to the arrival characteristics of data packets requested by the service, achieve delay QoS guarantee while minimizing resource occupancy; in addition, the method can also fully save network resources and achieve the same SFC carrying capacity with less consumption.

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

[0007] A VNF deployment method based on coupled bandwidth allocation and with latency QoS guarantee includes the following steps:

[0008] Step 1: Abstract the physical network, obtain the SFC to be deployed, and pre-allocate resources based on the characteristics of the arriving traffic based on queuing theory and large deviation theory.

[0009] Step 2: Sort the physical nodes in descending order according to the remaining CPU resource size, and select candidate physical nodes and candidate physical links according to the pre-allocated CPU resources and bandwidth resources, specifically:

[0010] Traverse the physical nodes in the physical network and sort them in descending order according to the remaining CPU resource size;

[0011] Select the physical nodes with the remaining CPU resources greater than the total CPU resources required by the VNF node and belonging to the server type as candidate nodes;

[0012] According to the order of SFC, two candidate nodes are selected as adjacent VNF ​​nodes in turn. If there is a direct physical link between the two candidate nodes and its bandwidth resources are greater than the effective bandwidth resource size pre-allocated by the virtual link, the physical link is used as the candidate link of this virtual link;

[0013] Step 3: Based on the martingale theory, estimate the coupling bandwidth after the new traffic is coupled with the existing traffic of the candidate link, and solve the coupling bandwidth increment of the candidate link;

[0014] Step 4: Solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy to obtain the optimal deployment solution; specifically:

[0015] The joint optimization problem includes: the objective function of the optimization problem is to minimize the network resource occupancy rate:

[0016]

[0017] Where: c ij represents the coupling bandwidth increment of the current request on the physical link (i, j), and The binary decision variables represent the matching relationship between physical nodes and virtual nodes, and the matching relationship between physical links and virtual links, respectively. Indicates that VNF ​​node u is deployed on physical node i. Indicates that u is not deployed on i. Indicates that the SFC virtual link (u, w) is deployed on the physical link (i, j). Indicates that (u, w) is not deployed on (i, j); each physical node i∈N s All of them have CPU resources, and their CPU resource capacity is , the current remaining CPU resources are ; Each physical link (i, j)∈Es has bandwidth resources, and its bandwidth resource capacity is , the current remaining bandwidth resource size is ; VNF node u∈N V The required CPU resource size is expressed as ;

[0018] The constraints of the optimization problem are:

[0019] Each VNF node must be deployed on and can only be deployed on one physical node:

[0020]

[0021] Traffic consistency constraints ensure that the physical nodes and physical links used as deployment solutions match. If a physical node is selected, the physical link connected to it must also be selected to ensure that the input traffic of the selected physical node is equal to the output traffic:

[0022]

[0023] Resource constraints constrain the CPU resource consumption of the deployed physical node to not exceed the CPU resource capacity of the current physical node:

[0024]

[0025] Resource constraints ensure that the bandwidth resource usage of the deployed physical link does not exceed the maximum capacity:

[0026]

[0027] Obtain current load information and solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy, that is, solve the decision variables and obtain the optimal deployment plan;

[0028] Update the physical nodes and physical link resources in the physical network topology according to the optimal deployment plan, check whether there is a resource overflow, if there is a resource overflow, solve the error, re-solve the current SFC, if there is no resource overflow, deploy the next SFC.

[0029] The content of step one is specifically as follows:

[0030] The physical network topology is abstracted as:

[0031] Abstract the physical network topology into a directed graph G s =(N s ,E s ), where N s represents the set of all physical nodes including switches and servers, E s Represents the set of links connecting all physical nodes in the topology;

[0032] In step 1, the physical network topology is abstracted as follows:

[0033] The physical network is represented as a directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers, E s Represents the set of physical links connecting all physical nodes in the topology;

[0034] The SFC to be deployed is expressed as:

[0035] SFC is divided into two layers of models. The first layer is the instance layer: SFC is represented as a directed graph G V =(N V , E V ), where N V represents the set of VNF nodes in SFC, E Vrepresents the set of all virtual links connecting VNF nodes; the second layer is the packet layer: each SFC carries a packet flow, each instance layer corresponds to a packet flow, and the packet flow is represented by e = {a e (t), D max , ε}, where a e (t) represents the number of packets arriving at the VNF node in time slot t from packet flow e, D max and ε are parameters that describe the delay requirements of packet flows. In packet flow e, the delay D e (t) needs to satisfy P{D e (t)≥D max}≤ε, homogeneous services have the same QoS requirements;

[0036] The resource pre-allocation is performed based on the traffic characteristics of the data packet flow arriving at the VNF node as follows:

[0037] Get the VNF nodes in each SFC, pre-allocate CPU resources according to their preset types, keep the CPU resources occupied by flow service-type VNF nodes in each SFC consistent, and randomly allocate the CPU resources occupied by packet service-type VNF nodes in each SFC differently;

[0038] According to the traffic characteristics of the data packet flow arriving at the VNF node on the virtual link in the SFC and its QoS requirements, the effective bandwidth on each virtual link is calculated using the effective bandwidth method to pre-allocate bandwidth resources;

[0039] The effective bandwidth of each virtual link is:

[0040] A multi-input and single-output queuing system is built on each link port of the physical server. When the data packet flow carried by the SFC passes through this port, the data to be processed will be sent to the corresponding queuing system and served according to the first-come-first-served principle. The queuing system at the port where server i is connected to server j is represented as S ij , the service rate is denoted as r ij ;

[0041] According to the arrival characteristics of each packet flow, the moment generating function M(θ) of the packet flow arriving at each VNF node is calculated, and the QoS requirement of the packet flow is expressed as P{D e (t)≥D max}≤ε, immediate delay D e (t) greater than D max The probability is less than ε;

[0042] According to the large deviation theory and Little's law, the delay violation probability of a constant rate queue system is: , where λ eis the average arrival rate of the packet flow, θ (θ ≥ 0) is the constraint index of the delay QoS;

[0043] According to queuing theory and large deviation theory, the effective bandwidth allocated on the virtual link (u, w) to meet the QoS requirements of the packet flow is calculated as where θ * is the solution of the QoS requirement of the packet flow under the delay violation probability of the queuing system, that is .

[0044] The content of step three is specifically as follows:

[0045] The traffic characteristics of multiple data flows arriving at the physical link port are obtained and integrated into a coupled flow. The statistical delay QoS requirement of the coupled flow is expressed as P{D(t)≥D max}≤ε, that is, the delay D(t) is greater than or equal to D max The probability that is less than or equal to ε;

[0046] The delay violation probability of the queuing system is calculated as:

[0047]

[0048] Where a(t) and s(t) represent the instantaneous values ​​of the number of arriving packets and the number of service packets, respectively, and h a (a(t)) is the relevant characteristic function of the process of coupling traffic reaching the physical link port, E[h a (a(0))] represents the characteristic function h related to the process of coupling traffic arriving at the physical link port a The expected value of (a(t)) at time slot 0, h s (s(t)) is the relevant characteristic function of the physical link port for the coupled traffic service process, E[h s (s(0))] represents the service process related characteristic function h s The expected value of (s(t)) at time slot 0, H is the threshold, indicating h a (a(t)) and h s The minimum value of the product (s(t)), φ * is a special martingale parameter that links the arrival martingale and service martingale parameters together, and is related to h a (a(t)) together describe the characteristics of the arrival process, K s represents the martingale correction parameter of the physical link port for the coupled traffic service process, and h s (s(t)) together describe the characteristics of the service process, and the above formula is solved to obtain

[0049]

[0050] Queuing Systemij The coupled bandwidth consumption of the top k flows is expressed as:

[0051]

[0052] in, It is specifically determined by the distribution of service processes in the queuing system;

[0053] The method of solving the coupling bandwidth increment of the candidate link is specifically:

[0054] The coupling bandwidth increment is defined as the coupling traffic bandwidth consumption after adding the current SFC minus the original load bandwidth, so the coupling bandwidth increment is expressed as:

[0055]

[0056] in, represents the coupling bandwidth increment of the kth flow on the physical link (i, j), represents the total coupled traffic bandwidth consumption of the kth flow, It represents the total coupled traffic bandwidth consumption after adding k-1 flows.

[0057] Beneficial effects of the present invention:

[0058] The present invention establishes a service function chain deployment model with QoS guarantee, which aims to minimize the utilization of network resources and jointly solves the two problems of QoS guarantee and resource conservation. The present invention scientifically and reasonably allocates limited physical resources in the network according to the traffic arrival characteristics of the service, effectively solving the QoS guarantee problem of the service; the coupled traffic bandwidth is estimated for homogeneous services in the slice, and the absolute bandwidth consumption of each request is converted into relative bandwidth consumption, realizing the coupled bandwidth allocation of the current load based on the martingale theory, effectively reducing the bandwidth consumption of the service, rationally utilizing network resources, and improving the network carrying capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of a scenario example to which the present invention can be applied;

[0060] Figure 2 It is a schematic diagram of a multiple-input single-output queuing system in the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0062] VNF (virtual network function, i.e. each virtual node) forms a virtual link, called SFC (service function chaining); QoS (quality of service).

[0063] A VNF deployment method based on coupled bandwidth allocation and with latency QoS guarantee includes the following steps:

[0064] Step 1: Abstract the physical network, obtain the SFC to be deployed, and pre-allocate resources based on the characteristics of the arriving traffic based on queuing theory and large deviation theory. Specifically:

[0065] The physical network topology is abstracted as:

[0066] Abstract the physical network topology into a directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers. Each physical server can host multiple VNFs, while the switch node only has the forwarding function. s Represents the set of links connecting all physical nodes in the topology; each physical node i∈N s All of them have CPU resources, and their CPU resource capacity is , the current remaining CPU resources are ; Each physical link (i, j)∈E s There are bandwidth resources on both of them, and their bandwidth resource capacity is , the current remaining bandwidth resource size is ;

[0067] In step 1, the physical network topology is abstracted as follows:

[0068] The physical network is represented as a directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers. Each physical server can host multiple VNFs, while the switch node only has the forwarding function. s Represents the set of links connecting all physical nodes in the topology; each physical node i∈N s All of them have CPU resources, and their CPU resource capacity is , the current remaining CPU resources are ; Each physical link (i, j)∈E sThere are bandwidth resources on both of them, and their bandwidth resource capacity is , the current remaining bandwidth resource size is ;

[0069] The SFC to be deployed is expressed as:

[0070] SFC is divided into two layers of models. The first layer is the instance layer: SFC is represented as a directed graph G V =(N V , E V ), where N V represents the set of VNF nodes in SFC, E V Represents the set of all virtual links connecting VNF nodes; there are two preset types of VNF nodes in SFC, namely, flow service type and packet service type, which have different resource allocation methods. VNF node u∈N V The required CPU resource size is expressed as ; Virtual link (u, w)∈E v The required bandwidth resource size is expressed as ;

[0071] The second layer is the packet layer: each SFC carries a packet flow, each instance layer corresponds to a packet flow, packets arrive at each time slot and the packets arriving at each time slot obey independent synchronization; the packet flow is represented by e = {a e (t), D max , ε}, where a e (t) represents the number of packets arriving at the VNF node in time slot t from packet flow e, D max and ε are parameters that describe the delay requirements of packet flows. In packet flow e, the delay D e (t) needs to satisfy P{D e (t)≥D max}≤ε, homogeneous services have the same QoS requirements;

[0072] The resource pre-allocation is performed based on the traffic characteristics of the data packet flow arriving at the VNF node as follows:

[0073] Get each VNF sequence, i.e., the VNF nodes in each SFC, and pre-allocate CPU resources according to their preset types. Keep the CPU resources occupied by the flow service-type VNF nodes in each SFC consistent, and randomly allocate the CPU resources occupied by the packet service-type VNF nodes in each SFC differently.

[0074] According to the traffic characteristics of the data packet flow arriving at the VNF node on the virtual link in the SFC and its requirements for quality of service (QoS), the effective bandwidth on each virtual link is calculated using the effective bandwidth method to pre-allocate bandwidth resources;

[0075] The effective bandwidth of each virtual link is:

[0076] A multi-input and single-output queuing system is built on each link port of the physical server. When the data packet flow carried by the SFC passes through this port, the data to be processed will be sent to the corresponding queuing system and served according to the first-come-first-served principle. The queuing system at the port where server i is connected to server j is represented as S ij , the service rate is denoted as r ij ;

[0077] According to the arrival characteristics of each packet flow, the moment generating function M(θ) of the packet flow arriving at each VNF node is calculated, and the QoS requirement of the packet flow is expressed as P{D e (t)≥D max}≤ε, immediate delay D e (t) greater than D max The probability is less than ε;

[0078] According to the large deviation theory and Little's law, the delay violation probability of a constant rate queue system is: where λ e is the average arrival rate of the packet flow, θ (θ ≥ 0) is the constraint index of the delay QoS;

[0079] Step 2: Sort the physical nodes in descending order according to the remaining CPU resource size, and select candidate physical nodes and candidate physical links according to the pre-allocated CPU resources and bandwidth resources, specifically:

[0080] Traverse the physical nodes in the physical network and sort them in descending order according to the remaining CPU resource size;

[0081] Select the physical nodes with the remaining CPU resources greater than the total CPU resources required by the VNF node and belonging to the server type as candidate nodes;

[0082] According to the order of SFC, two candidate nodes are selected as adjacent VNF ​​nodes in turn. If there is a direct physical link between the two candidate nodes and its bandwidth resources are greater than the effective bandwidth resource size pre-allocated by the virtual link, the physical link is used as the candidate link of this virtual link;

[0083] Step 3: Based on the martingale theory, estimate the coupling bandwidth after the new traffic is coupled with the existing traffic of the candidate link, and solve the coupling bandwidth increment of the candidate link, which is:

[0084] The coupling bandwidth is: the method considers that homogeneous services within a slice share the network bandwidth resources of the resource pool within the slice; when multiple data packet flows are carried on the same physical link, the total coupling bandwidth requirement is not equal to the direct superposition of the bandwidth requirements of the respective flows, but the flow characteristics of multiple flows are integrated into one coupled flow, and then allocated according to the coupled flow characteristics; the growth of the bandwidth does not follow a linear relationship, but is closely related to the current load, which is called coupling bandwidth.

[0085] The traffic characteristics of multiple data flows arriving at the physical link port are obtained and integrated into a coupled flow. The statistical delay QoS requirement of the coupled flow is expressed as P{D(t)≥D max}≤ε, that is, the delay D(t) is greater than or equal to D max The probability that is less than or equal to ε;

[0086] A(t) represents the number of data packets of coupled traffic arriving at the physical link port from time slot 0 to time slot t, and S(t) represents the number of data packets served by the physical link port from time slot 0 to time slot t;

[0087] The delay violation probability of the queuing system is calculated as:

[0088]

[0089] Where a(t) and s(t) represent the instantaneous values ​​of the number of arriving packets and the number of service packets, respectively, and h a (a(t)) is the relevant characteristic function of the process of coupling traffic reaching the physical link port, E[h a (a(0))] represents the characteristic function h related to the process of coupling traffic arriving at the physical link port a The expected value of (a(t)) at time slot 0, h s (s(t)) is the relevant characteristic function of the physical link port for the coupled traffic service process, E[h s (s(0))] represents the service process related characteristic function h s The expected value of (s(t)) at time slot 0, H is the threshold, indicating h a (a(t)) and h s The minimum value of the product (s(t)), φ * is a special martingale parameter that links the arrival martingale and service martingale parameters together, and is related to h a (a(t)) together describe the characteristics of the arrival process, K s represents the martingale correction parameter of the physical link port for the coupled traffic service process, and h s (s(t)) together describe the characteristics of the service process, and the above equation can be solved to obtain

[0090]

[0091] Queuing System ij The coupled bandwidth consumption of the top k flows is expressed as:

[0092]

[0093] in, It is specifically determined by the distribution of service processes in the queuing system;

[0094] The method of solving the coupling bandwidth increment of the candidate link is specifically:

[0095] The coupling bandwidth increment is defined as the coupling traffic bandwidth consumption after adding the current SFC minus the original load bandwidth, so the coupling bandwidth increment is expressed as:

[0096]

[0097] in, represents the coupling bandwidth increment of the kth flow on the physical link (i, j), represents the total coupled traffic bandwidth consumption of the kth flow, represents the total coupled traffic bandwidth consumption of k-1 flows;

[0098] Step 4: Solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy to obtain the optimal deployment solution; specifically:

[0099] The joint optimization problem includes:

[0100] The objective function of the optimization problem is to minimize the network resource occupancy rate:

[0101]

[0102] Where: c ij represents the coupling bandwidth increment of the current request on the physical link (i, j), and The binary decision variables represent the matching relationship between physical nodes and virtual nodes, and the matching relationship between physical links and virtual links, respectively. Indicates that VNF ​​node u is deployed on physical node i. Indicates that u is not deployed on i. Indicates that the SFC virtual link (u, w) is deployed on the physical link (i, j). Indicates that (u, w) is not deployed on (i, j);

[0103] In the model, the first part of formula (4) is to minimize the total CPU occupancy of the nodes, and the second part is to minimize the total bandwidth occupancy of the links in the network when different deployment decisions will result in different bandwidth consumption.

[0104] The constraints of the optimization problem are:

[0105] Each VNF node must be deployed on and can only be deployed on one physical node:

[0106]

[0107] Traffic consistency constraints ensure that the physical nodes and physical links used as deployment solutions match. If a physical node is selected, the physical link connected to it must also be selected to ensure that the input traffic of the selected physical node is equal to the output traffic:

[0108]

[0109] Resource constraints constrain the CPU resource consumption of the deployed physical node to not exceed the CPU resource capacity of the current physical node:

[0110]

[0111] Resource constraints ensure that the bandwidth resource usage of the deployed physical link does not exceed the maximum capacity:

[0112]

[0113] Obtain the current load information and solve the joint optimization problem of satisfying the latency QoS and the minimum resource occupancy rate, that is, solve the decision variables and obtain the optimal deployment plan. The current load information obtained is as follows:

[0114] For each SFC, obtain its candidate node and candidate link list;

[0115] Obtain the current load status of each candidate link, including the number of load flows and the arrival characteristics of load flows;

[0116] Calculate the coupling traffic bandwidth size after adding the current SFC on each candidate link, and calculate the coupling bandwidth increment c on each candidate link ij ;

[0117] Update the physical nodes and physical link resources in the physical network topology according to the optimal deployment plan, check whether there is a resource overflow, if there is a resource overflow, solve the error, re-solve the current SFC, if there is no resource overflow, deploy the next SFC.

[0118] Example 2

[0119] The embodiment of the present invention is based on NFV and SDN architecture, and proposes a VNF deployment method based on QoS guarantee and relative bandwidth growth. In the 5G network, diversified services are abstracted into service function chains SFC connected by multiple VNFs according to the types, scales and requirements of the services. These SFCs contain the types of virtual network functions required by the services and the size of network resources. Based on SFC, the network controller allocates resources to the services from the resource pool, plans physical paths, and implements customized network services.

[0120] Figure 1 The process of mapping SFC from virtual to physical network in the present invention is shown. There are two SFCs in the figure. Each SFC represents a service flow. Each SFC can contain different types of VNFs. Figure 1 There are two types of VNFs, distinguished by VNF a and VNF b. The SFC generated initially contains the VNF functional logical links of the service and the traffic arrival characteristics of the service. Next, after the first step of resource pre-allocation based on the traffic characteristics, the SFC contains the resource size required when it is deployed on the physical network, which are B1 and B2 respectively. Since the physical network of the resource pool already has an SFC that arrived first, it is regarded as a load, and it is assumed that the current load consumption is B0. When the SFC is deployed on the link in the physical network, the total resource consumption is not equal to the simple addition of the load resource consumption and the current SFC resource consumption, but the load bandwidth plus the coupled bandwidth increment of the two. Figure 1 In the example, part of SFC 1 is deployed on the link with load B0. The bandwidth consumed by this link is B0+c(B 0，1 ), where c(B 0，1 ) represents the coupling bandwidth increment between load 0 and SFC 1. The other part of SFC 1 and the part of SFC 2 are deployed together on the link of load B0. The bandwidth consumed by this link is B0+c(B 0，1，2 ), where c(B 0，1，2 ) is expressed as the coupling bandwidth increment of load 0, SFC 1 and SFC2.

[0121] In the embodiment of the present invention, the steps are as follows:

[0122] S101: Abstract the physical network, obtain the SFC to be deployed, and pre-allocate resources based on the characteristics of the arriving traffic based on the queuing theory and large deviation theory.

[0123] S102: sorting the physical nodes in descending order according to the remaining CPU resource size, and screening candidate physical nodes and candidate physical links according to the pre-allocated resources;

[0124] S103: Based on the martingale theory, estimate the coupling bandwidth after the new traffic is coupled with the existing traffic of the candidate link, and solve the coupling bandwidth increment of the candidate link;

[0125] S104: Solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy, and obtain the optimal deployment solution.

[0126] Further, in step S101, the physical network topology of the embodiment is abstracted, and the steps are:

[0127] S1011: Randomly generate basic topology of physical network based on Waxman model, represented as directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers. Each physical server can host multiple VNFs, while the switch node only has the forwarding function. s Represents the set of links connecting all nodes in the topology;

[0128] S1012: Allocate initial resources for the physical network, including the type of each physical node, the CPU resource capacity of the physical node, and the bandwidth resource capacity on the physical link.

[0129] Further, in step S101, the step of obtaining the SFC to be deployed is:

[0130] S1013: Acquire the arrival time and departure time of multiple preset SFCs, form a set, and check in each time slot whether there is an SFC arriving or leaving the network;

[0131] S1014: When the SFC reaches the network, the topology of the SFC is obtained, which is represented as a directed graph G. V =(N V , E V ), where N V represents the VNF set in SFC, E V Represents the set of all virtual links connecting VNFs; obtains the type of each virtual node VNF and the virtual link connection relationship;

[0132] S1015: Obtain the traffic characteristics of the data packets arriving at the SFC. In this embodiment, the arrival of the data packets follows the Poisson distribution as an example, and the traffic characteristics are the average arrival rate λ e , the QoS requirement of the obtained traffic is expressed as P{D e (t)≥D max}≤ε, immediate delay D e (t) greater than or equal to D max The probability is less than or equal to ε.

[0133] Further, in step S101, based on the queuing theory and the large deviation theory, resource pre-allocation is performed according to the flow characteristics of the arrival flow, and the steps are:

[0134] S1016: Obtain the VNF nodes in each SFC, pre-allocate CPU resources according to their preset types, keep the CPU resources occupied by the stream service type VNF in each VNF sequence consistent, and randomly allocate the CPU resources occupied by the packet service type VNF in each sequence differently;

[0135] S1017: According to the traffic characteristics of data packets arriving on the virtual links in the VNF sequence and their quality of service (QoS) requirements, based on queuing theory and large deviation theory, the effective bandwidth on each virtual link is calculated by the effective bandwidth method to pre-allocate bandwidth resources.

[0136] Further, in step S1017, the effective bandwidth method is specifically:

[0137] The arrival of data packets of the traffic in this embodiment follows a Poisson process, and the average arrival rate is λ e , QoS requirement is expressed as P{D e (t)≥D max}≤ε;

[0138] The moment generating function of a single flow Poisson process is M(θ)=exp(λ e (e θ -1)) According to queuing theory, large deviation theory and Little's law, the delay violation probability of a constant rate queue system is derived

[0139] Calculate the effective bandwidth that meets the traffic QoS requirements as where θ * is the solution of the QoS requirement of a single flow under the delay violation probability of the queuing system. In this embodiment

[0140] Further, in step S102, the step of screening candidate nodes and links is:

[0141] S1021: traverse the physical nodes in the physical network and sort them in descending order according to the remaining CPU size;

[0142] S1022: Select a physical node that is ranked first, has a remaining CPU resource greater than the resource size required by the VNF node, and is a server type as a candidate node;

[0143] S1023: Select candidate nodes of two adjacent VNF ​​nodes in turn according to the order of the VNF sequence. If there is a direct physical link between the two candidate nodes and the bandwidth resources are greater than the bandwidth resources pre-allocated by the virtual link, add the physical link to the candidate link of this virtual link.

[0144] Further, in step S103, the coupling bandwidth includes:

[0145] S1031: The method considers that homogeneous services within a slice share network bandwidth resources in a resource pool within the slice;

[0146] S1032: When multiple flows are carried on the same physical link, the total coupled bandwidth requirement is not equal to the direct addition of the bandwidth requirements of the individual flows. Instead, the flow characteristics of the multiple flows are integrated into one coupled flow, and then the coupled flow is allocated according to the coupled flow characteristics.

[0147] S1033: The growth of the bandwidth does not follow a linear relationship, but is closely related to the current load, which is called coupling bandwidth.

[0148] Further, in step S103, estimating the coupling bandwidth after coupling the new traffic with the existing traffic of the link based on the martingale theory includes:

[0149] S1034: Figure 2 Schematic diagram of a multi-input and single-output queuing system for coupled traffic. A queuing system is built on each link port to obtain the traffic characteristics of multiple arriving flows and integrate them into a coupled flow. The statistical delay QoS requirement of the coupled flow is expressed as P{D(t)≥D max}≤ε, that is, the delay D(t) is greater than or equal to D max The probability that is less than or equal to ε;

[0150] S1035: Solve the delay violation probability of the queuing system based on the martingale theory;

[0151] S1036: Estimation of coupled flow bandwidth based on martingale theory;

[0152] Further, in step S1035, solving the delay violation probability of the queuing system based on the martingale theory includes:

[0153] In this embodiment, the arrival of traffic packets follows the Poisson process. The coupled traffic is the superposition of multiple traffic flows, which still conforms to the distribution of the Poisson process. The average arrival rate is is the flow quantity, λ e Represents the average arrival rate of a single flow;

[0154] The data packet arrival process follows independent and identical distribution, h a =1 and

[0155] In this embodiment, the service process is a constant rate service, and the service rate is represented by C. s =1 and K s =C;

[0156] So we can get that the delay violation probability is

[0157]

[0158] Further, in step S1036, the coupling flow bandwidth estimation based on the martingale theory includes:

[0159] Queuing Systems ij The coupled bandwidth consumption of the last k flows is calculated as

[0160]

[0161] in, It is determined by the service process distribution of the queuing system.

[0162] Further, in step S103, the steps of solving the coupling bandwidth increment of the candidate link include:

[0163] S1038: define the coupling bandwidth increment as the coupling traffic bandwidth consumption after adding the current SFC minus the original load bandwidth. The coupling bandwidth increment is expressed by the following formula;

[0164]

[0165] In the above formula, represents the coupling bandwidth increment of the kth flow on the physical link (i, j), represents the total coupled traffic bandwidth consumption after adding k flows, Indicates the bandwidth consumption of the first k-1 flows.

[0166] Further, in step S104, the joint optimization problem of satisfying the delay QoS and the minimum resource occupancy rate comprises the following steps:

[0167] S1041: The objective function of the optimization model is to minimize the network resource occupancy rate:

[0168]

[0169] where c ij represents the coupling bandwidth increment of the current request on the physical link (i, j), and are binary decision variables representing the matching relationship between physical nodes and virtual nodes, and the matching relationship between physical links and virtual links, respectively;

[0170] S1042: The constraints of the optimization model are:

[0171]

[0172]

[0173]

[0174]

[0175] Further, in step S104, the specific steps are:

[0176] S1043: For each SFC, obtain a list of candidate nodes and candidate links;

[0177] S1044: Obtain the current load status of each candidate link, including the number of load flows and the arrival characteristics of the load flows;

[0178] S1045: Calculate the coupling traffic bandwidth size on each candidate link after adding the current VNF ​​sequence, and calculate the coupling bandwidth on each link;

[0179] S1046: Solve the decision variables and determine the deployment plan according to the model described in step S105;

[0180] S1047: Update the node and link resources in the physical network topology, check whether there is a resource overflow, if there is a resource overflow, solve the error, re-solve the current VNF ​​sequence, if there is no resource overflow, deploy the next VNF ​​sequence.

[0181] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A VNF deployment method based on coupled bandwidth allocation and with latency QoS guarantee, characterized in that The following steps are involved: Step 1: Abstract the physical network, obtain the SFC to be deployed, and pre-allocate resources based on the characteristics of the arriving traffic based on queuing theory and large deviation theory. Step 2: Sort the physical nodes in descending order according to the remaining CPU resource size, and select candidate physical nodes and candidate physical links according to the pre-allocated CPU resources and bandwidth resources, specifically: Traverse the physical nodes in the physical network and sort them in descending order according to the remaining CPU resource size; Select the physical nodes with the remaining CPU resources greater than the total CPU resources required by the VNF node and belonging to the server type as candidate nodes; According to the order of SFC, two candidate nodes are selected as adjacent VNF ​​nodes in turn. If there is a direct physical link between the two candidate nodes and its bandwidth resources are greater than the effective bandwidth resource size pre-allocated by the virtual link, the physical link is used as the candidate link of this virtual link; Step 3: Based on the martingale theory, estimate the coupling bandwidth after the new traffic is coupled with the existing traffic of the candidate link, and solve the coupling bandwidth increment of the candidate link; Step 4: Solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy to obtain the optimal deployment solution; specifically: The joint optimization problem includes: the objective function of the optimization problem is to minimize the network resource occupancy rate: Where: c ij represents the coupling bandwidth increment of the current request on the physical link (i, j), and The binary decision variables represent the matching relationship between physical nodes and virtual nodes, and the matching relationship between physical links and virtual links, respectively. Indicates that VNF ​​node u is deployed on physical node i. Indicates that u is not deployed on i. Indicates that the SFC virtual link (u, w) is deployed on the physical link (i, j). Indicates that (u, w) is not deployed on (i, j); each physical node i∈N s All of them have CPU resources, and their CPU resource capacity is The current remaining CPU resources are Each physical link There are bandwidth resources on both of them, and their bandwidth resource capacity is The current remaining bandwidth resource size is VNF node u∈N V The required CPU resource size is expressed as The constraints of the optimization problem are: Each VNF node must be deployed on and can only be deployed on one physical node: Traffic consistency constraints ensure that the physical nodes and physical links used as deployment solutions match. If a physical node is selected, the physical link connected to it must also be selected to ensure that the input traffic of the selected physical node is equal to the output traffic: Resource constraints constrain the CPU resource consumption of the deployed physical node to not exceed the CPU resource capacity of the current physical node: Resource constraints ensure that the bandwidth resource usage of the deployed physical link does not exceed the maximum capacity: Obtain current load information and solve the joint optimization problem of satisfying latency QoS and minimum resource occupancy, that is, solve the decision variables and obtain the optimal deployment plan; Update the physical nodes and physical link resources in the physical network topology according to the optimal deployment plan, check whether there is a resource overflow, if there is a resource overflow, solve the error, re-solve the current SFC, if there is no resource overflow, deploy the next SFC.

2. A VNF deployment method based on coupled bandwidth allocation and with latency QoS guarantee according to claim 1, characterized in that The content of step one is specifically as follows: The physical network topology is abstracted as: Abstract the physical network topology into a directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers, E s Represents the set of links connecting all physical nodes in the topology; In step 1, the physical network topology is abstracted as follows: The physical network is represented as a directed graph G s =(N s , E s ), where N s represents the set of all physical nodes including switches and servers, E s Represents the set of physical links connecting all physical nodes in the topology; The SFC to be deployed is expressed as: SFC is divided into two layers of models. The first layer is the instance layer: SFC is represented as a directed graph G V =(N V , E V ), where N V represents the set of VNF nodes in SFC, E V represents the set of all virtual links connecting VNF nodes; the second layer is the packet layer: each SFC carries a packet flow, each instance layer corresponds to a packet flow, and the packet flow is represented by e = {a e (t), D max , ε}, where a e (t) represents the number of packets arriving at the VNF node in time slot t from packet flow e, D max and ε are parameters that describe the delay requirements of packet flows. In packet flow e, the delay D e (t) needs to satisfy P{D e (t)≥D max }≤ε, homogeneous services have the same QoS requirements; The resource pre-allocation is performed based on the traffic characteristics of the data packet flow arriving at the VNF node as follows: Get the VNF nodes in each SFC, pre-allocate CPU resources according to their preset types, keep the CPU resources occupied by flow service-type VNF nodes in each SFC consistent, and randomly allocate the CPU resources occupied by packet service-type VNF nodes in each SFC differently; According to the traffic characteristics of the data packet flow arriving at the VNF node on the virtual link in the SFC and its QoS requirements, the effective bandwidth on each virtual link is calculated using the effective bandwidth method to pre-allocate bandwidth resources; The effective bandwidth of each virtual link is: A multi-input and single-output queuing system is built on each link port of the physical server. When the data packet flow carried by the SFC passes through this port, the data to be processed will be sent to the corresponding queuing system and served according to the first-come-first-served principle. The queuing system at the port where server i is connected to server j is represented as S ij , the service rate is denoted as r ij ; According to the arrival characteristics of each packet flow, the moment generating function M(θ) of the packet flow arriving at each VNF node is calculated, and the QoS requirement of the packet flow is expressed as P{D e (t)≥D max }≤ε, immediate delay D e (t) greater than D max The probability is less than ε; According to the large deviation theory and Little's law, the delay violation probability of a constant rate queue system is: where λ e is the average arrival rate of the packet flow, θ (θ ≥ 0) is the constraint index of the delay QoS; According to queuing theory and large deviation theory, the effective bandwidth allocated on the virtual link (u, w) to meet the QoS requirements of the packet flow is calculated as where θ * is the solution of the QoS requirement of the packet flow under the delay violation probability of the queuing system, that is 3. A VNF deployment method based on coupled bandwidth allocation and with latency QoS guarantee according to claim 2, characterized in that The content of step three is specifically as follows: The traffic characteristics of multiple data flows arriving at the physical link port are obtained and integrated into a coupled flow. The statistical delay QoS requirement of the coupled flow is expressed as P{D(t)≥D max }≤ε, that is, the delay D(t) is greater than or equal to D max The probability that is less than or equal to ε; The delay violation probability of the queuing system is calculated as: Where a(f) and s(t) represent the instantaneous values ​​of the number of arriving packets and the number of service packets, respectively, and h a (a(t)) is the relevant characteristic function of the process of coupling traffic reaching the physical link port, E[h a (a(0))] represents the characteristic function h related to the process of coupling traffic arriving at the physical link port a The expected value of (a(t)) at time slot 0, h s (s(t)) is the relevant characteristic function of the physical link port for the coupled traffic service process, E[h s (s(0))] represents the service process related characteristic function h s The expected value of (s(t)) at time slot 0, H is the threshold, indicating h a (a(t)) and h s The minimum value of the product (s(t)), φ * is a special martingale parameter that links the arrival martingale and service martingale parameters together, and is related to h a (a(t)) together describe the characteristics of the arrival process, K s represents the martingale correction parameter of the physical link port for the coupled traffic service process, and h s (s(t)) together describe the characteristics of the service process, and the above formula is solved to obtain Queuing System ij The coupled bandwidth consumption of the top k flows is expressed as: in, It is specifically determined by the distribution of service processes in the queuing system; The method of solving the coupling bandwidth increment of the candidate link is specifically: The coupling bandwidth increment is defined as the coupling traffic bandwidth consumption after adding the current SFC minus the original load bandwidth, so the coupling bandwidth increment is expressed as: in, represents the coupling bandwidth increment of the kth flow on the physical link (i, j), represents the total coupled traffic bandwidth consumption of the kth flow, It represents the total coupled traffic bandwidth consumption after adding k-1 flows.

Images