A QoS optimization method for SDN networks based on mimicry

Abstract

The present invention discloses a method for optimizing the QoS of an SDN network based on the concept of mimicry. By deploying redundant heterogeneous controllers, executors with higher credibility and higher network QoS index are selected to participate in the arbitration, and the arbitration flow table with better performance is output and sent to the data plane to complete the forwarding of traffic. At the same time, the credibility and QoS index of the heterogeneous controllers are dynamically adjusted based on the arbitration results of each executor and the real-time network performance measurement results to ensure that the reference features when selecting the executor are accurate and timely, thereby improving the accuracy and high quality of the arbitration results. This method can reduce the network paralysis problem caused by the security vulnerabilities of a single controller in the traditional SDN architecture, effectively improve the active defense capability of the SDN network, and greatly optimize the network performance of the SDN.

Description

Technical Field

[0001] The present invention belongs to the field of computer network technology, relates to the field of SDN network technology, and specifically relates to an SDN network QoS optimization method based on mimicry thinking. Background Art

[0002] Traditional network architectures have a strong coupling between routing, forwarding, and control. In today's diverse, complex, and demanding network transmission environment, such architectures are clearly inflexible and practical, and development costs are significantly increased. To address this dilemma, Software Defined Networking (SDN), a new network architecture that separates control from forwarding, has been proposed. The SDN architecture consists of three layers. The core layer is the control layer, which centrally controls the global network, processes network data, and issues flow tables. The control layer connects to the data forwarding layer and the application layer via southbound and northbound interfaces, respectively. The forwarding layer forwards traffic based on flow tables and collects network information. The application layer provides a rich set of external interfaces, allowing users to develop their own applications. Compared to traditional network architectures, SDN opens up network programming, reduces development and control costs, and significantly improves scalability and network performance. Consequently, SDN has attracted widespread attention from both academia and industry.

[0003] While SDN has brought significant efficiency and innovation to network architecture, significant security issues remain, particularly regarding the security of core controllers. In typical SDN architectures, controllers are deployed as single nodes. Attackers can exploit software and hardware vulnerabilities or launch malicious attacks, compromising the core layer. This prevents the forwarding layer from receiving forwarding measurements, potentially paralyzing the entire network. Therefore, research on enhancing the proactive defense capabilities of SDN architectures and improving the quality of service within SDN networks is crucial. Summary of the Invention

[0004] The purpose of the present invention is to address the deficiencies of the existing technology and provide an SDN network QoS optimization method based on mimicry thinking.

[0005] The object of the present invention is achieved through the following technical solution: a method for optimizing SDN network QoS based on mimicry, the method comprising the following steps:

[0006] (1) Construct a set of heterogeneous controller executive bodies E. The specific steps are as follows:

[0007] (1.1) Deploy M servers A={a i |i=1, 2,...,M}, where a i For the i-th server, a heterogeneous operating system is deployed;

[0008] (1.2) Deploy N microcontainers B on each server = {bj |j=1, 2, ..., N}, where b j For the jth microcontainer, in each microcontainer b j Heterogeneous operating systems, heterogeneous SDN controllers and other heterogeneous components are deployed on the system to construct heterogeneous controller executives. The set of heterogeneous controller executives is denoted as E = {e i |i=1,2,...,M*N},where e i is the i-th heterogeneous executive;

[0009] (2) Define the executor's decision feature value, introduce the credibility and QoS index, and use r i and s i Respectively represent the executive body e i The credibility and QoS index of the , and set the initial credibility value to r0;

[0010] (3) Select d executors to distribute traffic through the scheduling strategy, where d ≤ M*N. The specific steps are:

[0011] (3.1) If the system makes its first ruling, the credibility is 1 and the scheduling strategy is a random selection strategy;

[0012] (3.2) If the system is not the first decision, the scheduling strategy is the feature priority selection strategy, which first filters out the executors whose credibility is lower than the threshold μ, and then selects r from the remaining executors. i Gao Hes i High executive bodies participate in adjudication;

[0013] (4) The selected executors make decisions on the traffic flow and obtain the flow table result f i ;

[0014] (5) Based on f i Classify the decision bodies and count the outputs with the same f i The sum of the credibility of the adjudicator is recorded as the flow table credibility rf i , the highest rf i The flow table to which it belongs is the final decision result f w , and send it to the data plane;

[0015] (6) Adjust the credibility and QoS index based on the adjudication results and network quality. The specific steps are as follows:

[0016] (6.1) Compare the judgment results of each execution body f i and the final ruling w , when the results are consistent, do not adjust r i , if they are inconsistent, then r is reduced at the decay rate of η i , when it is lower than the threshold μ, the executor is offline and a backup executor is selected to go online;

[0017] (6.2) Use network measurement tools to test f w The current network throughput and latency indicators of the switch and system where the forwarding port is located, and the QoS index s of the execution body are calculated i ;At the same time, the index will be updated regularly during the operation of the system;

[0018] (7) The offline executor that performs the self-cleaning operation is cleaned, and isomerized. After setting the initial characteristic value, it is added to the standby executor set.

[0019] Furthermore, in step (2), the initial credibility value r0 is set according to the actual deployment of the system, and the QoS index is automatically set after the first ruling or the regular update policy.

[0020] Furthermore, in step (5), the adjudication bodies are classified by referring to whether the flow tables of their adjudications are the same. If they are the same, they are classified into one category; otherwise, they are classified into different categories.

[0021] Furthermore, in step (5), the flow table f is calculated according to formula (1) i Flow table credibility rf i :

[0022]

[0023] where r j The ruling result is f i The executive body j credibility.

[0024] Furthermore, in step (6.1), the credibility of the executive body is adjusted according to formula (2):

[0025] r′ i =r i -η*r i (2)

[0026] Among them, η is the credibility decay rate, r i is the credibility before attenuation, r′ i is the credibility after attenuation.

[0027] Furthermore, in the step (6.2), the measurement indicators include throughput, latency, maximum number of concurrent TCP connections and number of transactions per second.

[0028] Furthermore, in step (6.2), the QoS index s is calculated according to formula (3): i :

[0029] s i =ωtp *p tp (i)+ω rt *p rt (i)+ω mctcc *p mctcc (i)+ω tps *p tps (i) (3)

[0030] where p tp (i) p rt (i) p mctcc (i) p tps (i) represents throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second; ω tp 、ω rt 、ω mctcc 、ω tps They represent the calculation weights of the four measurement indicators: throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second.

[0031] Furthermore, in step (6.2), the QoS index s i There are two ways to update: one is feature adjustment after the ruling, and the other is scheduled update, where the time of scheduled update is set according to the system operation status.

[0032] Compared with existing technologies, the present invention has the following advantages: It deploys redundant, heterogeneous controllers based on mimicry to perform flow table arbitration, selects executors with high credibility and network QoS indexes for arbitration, outputs high-performance arbitration flow tables to the data plane to complete traffic forwarding, and dynamically adjusts the credibility and QoS indexes of heterogeneous executors in real time to ensure that the reference characteristics for executor selection are accurate and timely. This aspect can mitigate the problem of network paralysis caused by inherent vulnerabilities or threats in SDN single-controller architectures, effectively improve the SDN architecture's proactive defense capabilities, and significantly optimize the SDN network service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0034] The present invention will be described in detail below with reference to the accompanying drawings. Unless there is any conflict, the features of the following embodiments and implementations may be combined with each other.

[0035] The present invention provides an SDN network QoS optimization method based on the mimicry concept. By deploying redundant heterogeneous controllers to perform flow table arbitration, executors with higher credibility and higher network QoS index are selected to participate in the arbitration. The arbitration flow table with better output performance is sent to the data plane to complete traffic forwarding. The credibility and QoS index of the heterogeneous executors are adjusted in real time and dynamically to ensure that the reference characteristics of the selected executors are accurate and timely.

[0036] like Figure 1 As shown in the figure, the present invention provides an SDN network QoS optimization method based on the mimicry concept, which includes the following steps:

[0037] (1) Construct a set of heterogeneous controller executors E and deploy redundant and heterogeneous SDN controllers. The specific steps are as follows:

[0038] (1.1) Deploy M servers A={a i |i=1, 2,...,M}, where a i For the i-th server, a heterogeneous operating system is deployed;

[0039] (1.2) Deploy N microcontainers B on each server = {b j |j=1, 2, ..., N}, where b j For the jth microcontainer, in each microcontainer b j Heterogeneous operating systems, heterogeneous SDN controllers and other heterogeneous components are deployed on the system to build heterogeneous controller executives. The set of heterogeneous controller executives is denoted as E = {e i |i=1,2,...,M*N},where e i is the i-th heterogeneous executive;

[0040] (2) Define the executor decision feature value, introduce credibility to quantify the reliability of the executor decision result, introduce the network service quality (QoS) index to quantify the availability of the executor decision result, and use r i and s i Respectively represent the executive body e k The credibility and network QoS index of the network are calculated, and the initial credibility value is set to r0;

[0041] (3) The dynamic scheduling module selects d executors to distribute traffic through the following scheduling strategy, where d ≤ M*N. The specific steps are:

[0042] (3.1) If the system makes its first ruling, the credibility is 1 and the scheduling strategy is a random selection strategy;

[0043] (3.2) If the system is not the first decision, the scheduling strategy is the feature priority selection strategy, which first filters out the executors whose credibility is lower than the threshold μ, and then selects r from the remaining executors.i Gao Hes i High executive bodies participate in adjudication;

[0044] (4) The selected executors make decisions on the traffic flow and send the flow table results f i Sent to the mimicry adjudication module;

[0045] (5) The decision module is based on f i Classify the decision bodies and count the outputs with the same f i The sum of the credibility of the adjudicator is recorded as the flow table credibility rf i , the highest rf i The flow table to which it belongs is the final decision result f w , and send it to the data plane;

[0046] (6) The feature adjustment module adjusts the credibility and QoS index according to the adjudication results and network quality. The specific steps are as follows:

[0047] (6.1) Compare the judgment results of each execution body f i and the final ruling w , when the results are consistent, do not adjust r i , if they are inconsistent, then r is reduced at the decay rate of η i , when it is lower than the threshold μ, the executor is offline and a backup executor is selected to go online;

[0048] (6.2) Use network measurement tools to test f w The current network throughput and latency of the switch and system where the forwarding port is located are used to calculate the QoS index s of the execution body. i ;At the same time, the index will be updated regularly during the operation of the system;

[0049] (7) The self-cleaning module cleans the offline executor that performs the self-cleaning operation, performs isomerization processing, sets the initial characteristic value, and adds it to the standby executor set.

[0050] Furthermore, in step (2), the initial value of the credibility r0 can be set according to the actual deployment of the system, and is generally 1. The QoS index does not have an initial value at this time, and is automatically set by the feature adjustment module after the first decision or the regular update of the policy. Furthermore, in step (5), the decision bodies are classified by referring to whether the flow table entries they decide are the same. If they are the same, they are classified into the same category; otherwise, they are classified into different categories.

[0051] Furthermore, in step (5), the flow table f is calculated according to formula (1) i The flow table reliability rf i :

[0052]

[0053] where r j The ruling result is f i The executive body j Credibility, and L≤d.

[0054] Furthermore, in step (6.1), the credibility of the executive body is adjusted according to formula (2):

[0055] r′ i =r i -η*r i (2)

[0056] Among them, η is the credibility decay rate, r i is the credibility before attenuation, r′ i is the credibility after attenuation.

[0057] Furthermore, in the step (6.2), the measurement indicators include throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second.

[0058] Furthermore, in step (6.2), the QoS index s is calculated according to formula (3): i :

[0059] s i =ω tp *p tp (i)+ω rt *p rt (i)+ω mctcc *P mctcc (i)+ω tp s*p tp s(i) (3)

[0060] where p tp (i) p rt (i) p mctcc (i) p tps (i) represents the executive body e i The output flow table corresponds to the switch and network measurement indicators, namely throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second, which are measured, summarized, and normalized values; ω tp 、ω rt 、ω mctcc 、ω tps They represent the calculation weights of the four measurement indicators: throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second. The specific values ​​are determined according to the actual application situation and meet the requirements of ω. tp +ω rt +ω mctcc +ω tps=1.

[0061] Furthermore, in step (6.2), the QoS index s i There are two ways to update, one is feature adjustment after judgment, the other is scheduled update. The time of scheduled update is set according to the system operation. If the system is active, that is, there is traffic inflow within 10 minutes, then the scheduled update time is not set. If it is inactive, the update time is set to 10 minutes, that is, the system collects relevant data every 10 minutes, processes it and updates it. i .

[0062] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0063] The above embodiments are intended only to illustrate the design concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. The scope of protection of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design concepts disclosed in the present invention are within the scope of protection of the present invention.

Claims

1. A SDN network QoS optimization method based on mimicry, characterized in that: The steps include: (1) Construct a set of heterogeneous controller executive bodies E. The specific steps are as follows: (1.1) Deploy M servers A={a i |i=1,2,...,M}, where a i For the i-th server, a heterogeneous operating system is deployed; (1.2) Deploy N microcontainers B on each server = {b j |j=1,2,...,N}, where b j For the jth microcontainer, in each microcontainer b j Heterogeneous operating systems, heterogeneous SDN controllers and other heterogeneous components are deployed on the system to construct heterogeneous controller executives. The set of heterogeneous controller executives is denoted as E = {e i |i=1,2,...,M*N}, where e i is the i-th heterogeneous executive; (2) Define the executor's decision feature value, introduce the credibility and QoS index, and use r i and s i Respectively represent the executive body e i The credibility and QoS index of the , and set the initial credibility value to r0; (3) Select d executors to distribute traffic through the scheduling strategy, where d ≤ M*N. The specific steps are: (3.1) If the system makes its first ruling, the credibility is 1 and the scheduling strategy is a random selection strategy; (3.2) If the system is not the first decision, the scheduling strategy is the feature priority selection strategy, which first filters out the executors whose credibility is lower than the threshold μ, and then selects r from the remaining executors. i Gao Hes i High executive bodies participate in adjudication; (4) The selected executors make decisions on the traffic flow and obtain the flow table result f i ; (5) Based on f i Classify the decision bodies and count the outputs with the same f i The sum of the credibility of the adjudicator is recorded as the flow table credibility rf i , the highest rf i The flow table to which it belongs is the final decision result f w , and send it to the data plane; (6) Adjust the credibility and QoS index based on the adjudication results and network quality. The specific steps are as follows: (6.1) Compare the judgment results of each execution body f i and the final ruling w , when the results are consistent, do not adjust r i , if they are inconsistent, then r is reduced at the decay rate of η i , when it is lower than the threshold μ, the executor is offline and a backup executor is selected to go online; (6.2) Use network measurement tools to test f w The current network throughput and latency indicators of the switch and system where the forwarding port is located, and the QoS index s of the execution body are calculated i ;At the same time, the index will be updated regularly during the operation of the system; (7) The offline executor that performs the self-cleaning operation is cleaned, and isomerized. After setting the initial characteristic value, it is added to the standby executor set.

2. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In the step (2), the initial credibility value r0 is set according to the actual deployment of the system, and the QoS index is automatically set after the first ruling or the regular update policy.

3. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In the step (5), the adjudication bodies are classified by referring to whether the flow tables of their adjudications are the same. If they are the same, they are classified into one category; otherwise, they are classified into different categories.

4. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In step (5), the flow table f is calculated according to formula (1) i The flow table reliability rf i : where r j The ruling result is f i The executive body j credibility.

5. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In step (6.1), the credibility of the executive body is adjusted according to formula (2): r′ i =r i -h*r i (2) Among them, η is the credibility decay rate, r i is the credibility before attenuation, r′ i is the credibility after attenuation.

6. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In the step (6.2), the measurement indicators include throughput, latency, maximum number of concurrent TCP connections and number of transactions per second.

7. The SDN network QoS optimization method based on mimicry concept as claimed in claim 6, characterized in that: In step (6.2), the QoS index s is calculated according to formula (3): i : s i =ω tp *p tp (i)+ω rt *p rt (i)+ω mctcc *p mctcc (i)+ω tps *p tps (i) (3) where p tp (i) p rt (i) p mctcc (i) p tps (i) represents throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second; ω tp 、ω rt 、ω mctcc 、ω tps They represent the calculation weights of the four measurement indicators: throughput, latency, maximum number of concurrent TCP connections, and number of transactions per second.

8. The SDN network QoS optimization method based on mimicry concept as claimed in claim 1, characterized in that: In the step (6.2), the QoS index s i There are two ways to update: one is feature adjustment after the ruling, and the other is scheduled update, where the time of scheduled update is set according to the system operation status.

Images