CN109286528B - SDN network multi-controller deployment method based on time delay - Google Patents

Abstract

The invention discloses a time delay-based SDN network multi-controller deployment method, which comprises the steps of firstly analyzing time delay and structure of network topology, calculating time delay between every two switches and the degree of every node in a network, determining the number of the switches which can be deployed by every controller, then selecting a node which meets conditions according to a custom algorithm as a deployment position of the controller, determining all the switches in a management area, and obtaining the minimum average time delay size and the minimum time delay size under the worst condition according to a minimum average time delay model and the minimum time delay model under the worst condition; calculating the residual nodes meeting the conditions by the same method to be used as the deployment position of another controller to obtain the minimum average time delay and the minimum time delay under the worst condition; the above process is repeated until the controller position is deployed. The method of the invention can be suitable for multi-controller position deployment algorithms with various network scales, and can solve the problem of controller position deployment systematically.

Description

SDN network multi-controller deployment method based on time delay

Technical Field

The invention relates to the technical field of computer networks, in particular to a time delay-based SDN network multi-controller deployment method.

Background

SDN creatively separates the control plane of network devices from the data forwarding functions of the devices and fully centralizes the control functions on the SDN controller. Different from the distributed control of the traditional network, the centralized control of the SDN enables the control plane to obtain the global information of network resources, so that the optimization and allocation of the resources are realized according to the requirements. Meanwhile, the network equipment is logically integrated through centralized control of the network, field configuration and maintenance of each equipment are not needed, and convenience of network control and maintenance is greatly improved. In order to improve the performance and robustness of the network, a mechanism that a plurality of controllers work simultaneously can be used during network deployment, the plurality of controllers manage the same switch, and when one controller fails, a standby controller can guarantee that the network continues to work normally. Meanwhile, the multi-controller divides the network into a plurality of areas, and each controller manages one of the areas. Thus, the load of the controller is reduced, and the expandability of the control plane is greatly improved.

While providing a number of benefits, SDN multi-controller deployment also provides a new problem, namely how many controllers need to be placed in a given SDN network topology, where each controller should be placed, and which switches each controller should manage. Different deployment positions of the controller have different influences on network overhead, the existing technology pays attention to performance to neglect the overhead, and the spanning type progress is achieved in small-sized network deployment, but the existing technology is greatly limited in large-sized or even medium-sized network deployment, namely, the cost overhead is too large, and partial areas are further optimized, so that the existing controller deployment scheme has great limitation.

Disclosure of Invention

The invention aims to overcome the defects in the background technology, and provides a time-delay-based SDN network multi-controller deployment method, which comprises the steps of performing optimization control on the position deployment of SDN multi-controllers, performing controller position optimization modeling by taking time delay as a standard, establishing a minimum average time delay model and a worst-case minimum time delay model, and finally, on the basis of a greedy algorithm and a shortest routing algorithm, taking the time delay model as a constraint condition, so that the method can be suitable for multi-controller position deployment algorithms of various network scales, and can solve the controller position deployment problem systematically.

In order to achieve the technical effects, the invention adopts the following technical scheme:

a SDN network multi-controller deployment method based on time delay comprises the following steps:

A. analyzing the time delay and the structure of the network topology, calculating the time delay between each switch and the degree of each node in the network (wherein the degree of the node is the branch of the node, and the degree of the node is several if there are several branches), and determining the number of switches which can be deployed by each controller;

B. selecting a node meeting the conditions as a deployment position of the controller according to a custom algorithm, and determining all switches in a management area of the controller; obtaining the minimum average time delay size and the minimum time delay size under the worst condition according to the minimum average time delay model and the minimum time delay model under the worst condition;

C. and repeating the steps until the deployment of the controller position is completed.

According to the technical scheme, the number and the deployment position of the controllers and the switches to be managed by each controller can be determined according to the specific technical scheme in the deployment of the SDN multiple controllers for a given SDN network topology, so that the utilization rate of the network topology is the highest, and the worst-case delay of the average delay domain of the whole network is ensured.

Further, the step B specifically includes the following steps:

B1. calculating the degree of each node, and selecting the first m nodes as the placement positions of the alternative controllers according to descending order; wherein m is the number of controllers in the network;

B2. selecting any one of the m alternative controller placing positions obtained in the step B1 as a controller placing position;

B3. calculating the shortest path from each switch to the controller of the selected placement position according to a routing algorithm;

B4. obtaining a minimum average delay value and a minimum delay value under the worst condition according to the minimum average delay model and the minimum delay model under the worst condition;

B5. respectively arranging the minimum average time delay value and the minimum time delay value under the worst condition in an ascending order, and selecting the first q-1 switches and the current controller to be merged into a control domain, wherein q is the number of switches manageable by each controller;

B6. and selecting another placement position from the alternative controller placement positions in the B1, and repeating the steps B3 to B5, so that the network is divided into m control domains, wherein each control domain comprises q nodes.

Further, the SDN network multi-controller deployment method based on the time delay specifically includes the following steps:

step 1: for network topology G ═ (V, E), V ═ V₁,v₂,...,v_nAnalyzing the time delay and the structure, determining the time delay among all the switches, the number m of the controllers and the number q of the switches manageable by each controller, and initializing a controller set C ═ C₁,c₂,...,c_mIn (c) } the reaction solution is,

where G ═ (V, E) denotes a network topology, V ═ V₁,v₂,...,v_nDenotes a set of n nodes of the network topology, E is a set of edges, n is a total number of switches, C ═ C₁,c₂,...,c_mDenotes the set of controllers in the network topology;

step 2: calculating the degree of each node, taking the degree of each node as a quality standard, arranging the nodes in descending order of the degree, selecting the first m nodes as alternative controller placement positions, and adding a controller set C ═ C₁,c₂,...,c_mIn (1) };

and step 3: traversing all switches, calculating each network node to controller c using a routing algorithm₁And f of each switch is calculated according to the obtained path_mean(v, c) and f_min(v, c), and for f_mean(v, c) and f_min(v, c) performing ascending order arrangement, and selecting the first q-1 switches and the controller c in the order_iIncorporating control Domain C_iFinally, deleting q nodes in the selected control domain from the network;

wherein f is_mean(v, c) is the minimum average delay value, f_min(v, c) a worst case minimum delay value;

and 4, step 4: choose C ═ C₁,c₂,...,c_mThe next controller in the row repeats the operation of step 3;

and 5: according to the steps, after the traversal of the m controllers is completed, a controller deployment scheme based on the minimum average time delay and the worst case minimum time delay can be obtained.

Compared with the prior art, the invention has the following beneficial effects:

the SDN network multi-controller deployment method based on time delay is based on a greedy algorithm, the degree of nodes is used as a quality standard, alternative controllers are selected firstly, then a routing algorithm is used for calculating the shortest path from each network node to the controller, the minimum average time delay (the maximum minimum time delay) is calculated according to the required path, the minimum average time delay (the maximum minimum time delay) is arranged in an ascending order (a descending order), the time delay is used as the basis of network performance, the time delay of each process of receiving and processing information flow in a deployment network is modeled and analyzed by a queuing theory method, a mathematical expression of the average time delay and the minimum time delay is obtained, the quantity, the deployment position and the switch to be managed by each controller can be determined by a specific technical scheme for a given SDN network topology in the deployment of the SDN multi-controller, so that the utilization rate of the network topology is the highest, the method ensures the minimum delay of the average delay domain of the whole network under the worst condition, can be suitable for the algorithm of multi-controller position deployment of various network scales, and solves the problem of controller position deployment systematically.

Drawings

Fig. 1 is a schematic diagram of a controller queuing system model in the delay-based SDN network multi-controller deployment method of the present invention.

Fig. 2 is a schematic diagram of a queuing model of data streams arriving at a controller in the latency-based SDN network multi-controller deployment method of the present invention.

FIG. 3 is a schematic diagram of a multi-controller network topology deployment optimization in an embodiment of the invention.

Detailed Description

The invention will be further elucidated and described with reference to the embodiments of the invention described hereinafter.

Example (b):

the first embodiment is as follows:

the technical scheme disclosed by the embodiment mainly aims to solve the problem of how to determine the number of controllers, the deployment positions and the switches to be managed by each controller through an explicit algorithm for a given SDN network topology in the deployment of the SDN multi-controller so as to maximize the utilization rate of the network topology.

To specifically explain the technical solution, first, a multi-controller deployment scheme in an ideal state needs to be introduced, which is also a technical solution commonly adopted in the prior art:

in an ideal state, the multi-controller deployment scheme measures the position deployment of the multi-controllers by two indexes of average time delay and worst-case time delay, but a plurality of time delay processes are ignored in a rational mode, which is the most common defect existing in the prior art;

specifically, the calculation formula based on the average delay index value is as follows:

in the model, the network is abstracted into a weighted graph, V represents a set of nodes in the network, S' represents a path set of the network, the time delay among the nodes is the weight of an edge in the weighted graph, min d (V, S) represents the shortest path from the node V to the node S, and n is the total number of the nodes.

Specifically, the worst propagation delay, i.e., the maximum value of the minimum delays, is calculated by the following formula:

the worst-case delay refers to the maximum in the shortest path from each node to the controller-deployed node.

In practical applications, the above scheme is problematic because in a wide area network, propagation delay is the most dominant factor for generating delay, and thus, the delay index in the above scheme only considers propagation delay. In a general network, the transmission delay and the controller delay are non-negligible factors.

The technical scheme of the implementation is as follows:

firstly, analyzing the time delay and the structure of the network topology, calculating the time delay between each switch and the degree of each node in the network, determining the number of switches which can be deployed by each controller,

then selecting a node meeting the conditions as a deployment position of the controller according to a custom algorithm, determining all switches in a management area of the controller, and obtaining a minimum average delay size and a worst minimum delay size according to a minimum average delay model and a worst minimum delay model;

calculating the residual nodes meeting the conditions by the same method to be used as the deployment position of another controller to obtain the minimum average time delay and the minimum time delay under the worst condition;

the above process is repeated until the controller position is deployed.

Specifically, the network topology G ═ (V, E), V ═ V₁,v₂,...,v_nAnd concretely describing a specific implementation mode of realizing the optimal deployment scheme of the controller by using the autonomous development algorithm by taking the example as follows:

step 1: determining n switches in the network topology, delay or length among links, the number m of controllers, the number q of switches manageable by each control, and an initialization controller set C { [ C { [₁,c₂,...,c_mIn (1) };

step 2: calculating the degree of each node, taking the degree of each node as a quality standard, arranging the nodes in descending order, selecting the first m nodes as alternative controllers, and adding a controller set C ═ C₁,c₂,...,c_mIn (1) };

and step 3: traversing all switches, calculating each network node to controller c using a routing algorithm₁And f is calculated according to the obtained paths_mean(v, c) and f_min(v, c), then on f_mean(v, c) and f_min(v, c) performing ascending order arrangement, and selecting the first q-1 switches and the controller c in the order_iIncorporating control Domain C_iIn the end, the selected control domain C_iQ nodes in (2) are deleted from the network;

and 4, step 4: choose C ═ C₁,c₂,...,c_mThe next controller in the row repeats the operation of step 3;

and 5: according to the steps, after the traversal of the m controllers is completed, a controller deployment scheme based on the minimum average time delay and the worst case minimum time delay can be obtained.

For better understanding of the above technical solution, the following explanation is also needed:

firstly, controller modeling based on queuing theory:

as shown in fig. 1 and fig. 2, which are a queuing system model and a queuing model of data stream arriving at a controller, respectively, wherein processing delay and waiting delay of the controller can be obtained through the queuing system model, where λ is an arrival rate of the data stream, and μ is an average service time of the controller; the queuing model of the data stream to the controller is the process of representing the data stream to the controller, and can be established as M^kQueuing procedure of/M/1, where λ_iIs the rate at which the ith piece of data arrives at the controller, satisfying λ₁+λ₂+.....+λ_k＝kλ。

In the above model, before the controller time delay can be calculated, the system should be set as follows:

(1) the controller accesses a plurality of switches. Before each switch is accessed, an establishment request is generated;

(2) the arrival of the data stream is random, and the Poisson distribution is met;

(3) the controller processes the data flow randomly and independently, and negative exponential distribution with average service time of 1/mu is satisfied;

(4) the buffer amount of the controller is infinite.

The process of data flow to the controller can be established as M based on the assumption that it has been made^kQueuing process of/M/1; according to M^kthe/M/1 system solves the delay generated by the controller, and supposing that n data flows are served in the system when the k data flow arrives at the server, the k data flow is in a waiting state in the period.

Then the average service time of each data stream when the K data streams are served in the service system is:

the average residence time of the data stream within the controller is:

can be obtained by finishing

I.e. controller delay T_cComprises the following steps:

in a large-scale network, the distance between network nodes is very long, so that the transmission of information between network links will also generate non-negligible delay, and in addition, the delay generated by the switch is the propagation delay and transmission delay.

Specifically, the propagation delay depends on the length between network links and the propagation speed of information, and the propagation delay expression is as follows: t is_p＝d(v,c)/R

Where d (v, c) represents the shortest distance between the switch v and the controller c, and R represents the propagation speed of the network link.

The average propagation delay and the minimum propagation delay obtained from the above equation are:

T_p-min＝mind(v,c)/R

specifically, the transmission delay depends on the data volume of the transmitted data packet, the transceiving rate of the switch interface and the number of nodes on a link through which information is received from the transmitted data to the controller, and the mathematical expression of the transmission delay is as follows:

T_t＝[(N(v,c)+1)×I(v,c)]/S

where N (v, c) represents the number of nodes between the switch and the controller, I (v, c) represents the amount of data to be transferred, and S represents the transceiving rate of the switch port.

The average transmission delay and the minimum transmission delay obtained from the above formula are respectively:

T_t-min＝min[(N(v,c)+1)×I(v,c)]/S

thus, a final delay model can be obtained, specifically including a minimum average delay model and a minimum delay model under the worst condition.

Wherein the minimum average delay model is

Namely:

namely:

the worst case minimum delay model is maxmin (T)_c+T_p+T_t) Namely:

namely: f. of_min(v,c)＝T_c+T_p-min+T_t-min

Example two

As shown in fig. 3, when there are multiple controllers in the network topology, each controller manages switches in a certain area, such as 5 switches in fig. 3, which can be divided into two areas, and the switches in each area can only be controlled by one controller, so that the partitions have five cases, namely, an AB-CDE area, a BC-ADE area, a CD-ABE area, a BD-ACE area, and a CE-ABD area.

The optimal arrangement scheme in each control area is the same as the arrangement scheme of a single controller, and the steps are as follows:

s1, when an AB-CDE area is divided, respectively calculating a minimum average time delay scheme in an AB area and a minimum average time delay scheme in a CDE area;

s2, respectively calculating the minimum average time delay scheme by using the other four partitions in the same manner as the first step;

and S3, comparing the minimum average time delay of the five partitions, wherein the minimum average time delay is the optimal scheme.

In this embodiment, the following is calculated by using the technical solution of the first embodiment: the controller in the area AB should be placed at the position of the switch B, the controller in the area CDE should be placed at the position of the switch C according to the calculation of the single-controller deployment scheme.

Calculating the other four partition conditions in the same way, and finally comparing the controller placement schemes when the AB-CDE partition can be achieved to meet the minimum average time delay model; the controller placement scheme when partitioned with CE-ABD meets the worst case minimum latency, the controller is at switch D, C.

It will be understood that the above embodiments are merely exemplary embodiments taken to illustrate the principles of the present invention, which is not limited thereto. It will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit and substance of the invention, and these modifications and improvements are also considered to be within the scope of the invention.

Claims (2)

1. A SDN network multi-controller deployment method based on time delay is characterized by comprising the following steps:

A. analyzing the time delay and the structure of the network topology, calculating the time delay between each switch in the network, and determining the number of the switches which can be deployed by each controller;

B. selecting a node meeting the conditions as a deployment position of the controller according to a custom algorithm, and determining all switches in a management area of the controller; obtaining a minimum average delay value and a minimum delay value under the worst condition according to the minimum average delay model and the minimum delay model under the worst condition; wherein the minimum average delay model is

The worst case minimum delay model is max min (T)_c+T_p+T_t)；T_cA controller time delay; t is_pIs the propagation delay; propagation delay depends on the length between network linksAnd the propagation speed of the information; t is_tIs the transmission delay; the transmission delay depends on the data volume of the transmitted data packet, the transceiving rate of the switch interface and the number of nodes on the link through which the information is received from the transmitted data to the controller;

the step B specifically comprises the following steps:

B1. calculating the degree of each node, and selecting the first m nodes as the placement positions of the alternative controllers according to descending order; wherein m is the number of controllers in the network;

B2. selecting any one of the m alternative controller placing positions obtained in the step B1 as a controller placing position;

B3. calculating the shortest path from each switch to the controller of the selected placement position according to a routing algorithm;

B4. obtaining a minimum average delay value and a minimum delay value under the worst condition according to the minimum average delay model and the minimum delay model under the worst condition;

B5. respectively arranging the minimum average time delay value and the minimum time delay value under the worst condition in an ascending order, and selecting the first q-1 switches and the current controller to be merged into a control domain, wherein q is the number of switches manageable by each controller;

B6. selecting another placement position from the placement positions of the alternative controllers in the B1, and repeating the steps B3 to B5, so that the network is divided into m control domains, and each control domain has q nodes;

C. until the controller position is deployed.

2. The delay-based SDN network multi-controller deployment method of claim 1, wherein the delay-based SDN network multi-controller deployment method specifically comprises the following steps:

step 1: for network topology G ═ (V, E), V ═ V₁，v₂，...，v_nAnalyzing the time delay and the structure, determining the time delay among all the switches, the number m of the controllers and the number q of the switches manageable by each controller, and initializing a controller set C (C)₁,c₂,...,c_m}，

Where G ═ (V, E) denotes a network topology, V ═ V₁，v₂，...，v_nDenotes the set of n nodes of the network topology, E is the set of edges, n is the total number of switches, C ═ C₁,c₂,...,c_mDenotes the set of controllers in the network topology;

step 2: calculating the degree of each node, taking the degree of each node as a quality standard, arranging the nodes in descending order of the degree, selecting the first m nodes as alternative controller placement positions, and adding a controller set C ═ C₁,c₂,...,c_mIn (1) };

and step 3: traversing all switches, calculating each network node to controller c using a routing algorithm₁And f of each switch is calculated according to the obtained path_mean(v, c) and f_min(v, c), and for f_mean(v, c) and f_min(v, c) performing ascending order arrangement, and selecting the first q-1 switches and the controller c in the order₁Incorporating control Domain C₁Finally, deleting q nodes in the selected control domain from the network;

wherein f is_mean(v, c) is the minimum average delay value, f_min(v, c) is the worst case minimum delay value;

and 4, step 4: choose C ═ C₁,c₂,...,c_mThe next controller in the row repeats the operation of step 3;

and 5: according to the steps, after the traversal of the m controllers is completed, a controller deployment scheme based on the minimum average time delay and the worst case minimum time delay can be obtained.

Images