CN106488303A - A kind of net cast network transmission performance optimization method based on software definition and system - Google Patents

Abstract

The invention provides a kind of net cast network transmission performance optimization method based on software definition and system, the present invention determines sample data by obtaining global network link operation information, and according to the sample data for determining, calculating network link congestion guide price；Current network link congestion state is obtained, and according to network link congestion guide price and the current network link congestion state, determines current network link congestion price vector；The current network link congestion price vector is sent to network information sending port, and the current network link congestion price vector received according to the network information sending port, adjust the transmission rate of the network information sending port, avoid network congestion, solution high speed network transmission performance optimization problem is found up from internal cause, improve the data transmission performance of net cast network.

Description

A method and system for optimizing network transmission performance of live video based on software definition

technical field

The invention relates to the technical field of live video network transmission, in particular to a method and system for optimizing the performance of live video network transmission based on software definition.

Background technique

With the continuous development of Internet technology, the scope of live video broadcasting is becoming wider and wider, such as press conferences, sports competitions, live teaching exchanges, business promotion, teleconferences, school opening ceremonies, celebrations, wedding ceremonies, etc. Network audio-visual has become It defines the main network behaviors of current network users, and it is also a must for major portals to strive for user profits. The existing network mainly uses CDN, P2P, transparent proxy technology and other solutions to improve the quality of video transmission. The above method can ensure the smooth playback of on-demand videos, but cannot guarantee the smooth playback of live videos.

In order to improve network service quality, software-defined networking (Software Defined Networking, SDN) technology emerges as the times require. Current research on software-defined networking mainly considers flow control from the perspective of network traffic load balancing, that is, constantly updating the switch flow table and forwarding traffic to links with light loads. But when all the switch links in the network are fully loaded, the method of updating the flow table is useless. Therefore, it is still necessary to find a way from the internal factors of the high-speed network system. By increasing the price of the link, suppressing the sending traffic of the source, and then combining it with flow control, it is a good way to really improve the transmission performance of video network live broadcast.

The FAST TCP protocol (FAST protocol for short) actively adjusts the sending rate according to the network congestion status at the sending end, actively avoids buffer queue overflow and congestion, and achieves higher link utilization and stability. It belongs to the high-speed network system However, the FAST protocol has the defect that the sender cannot accurately obtain the network congestion status. The explicit congestion notification protocol (ECN protocol for short) can notify the sender of the network congestion status, but the ECN protocol can only notify the local link congestion status. Therefore, there is an urgent need for a method to effectively combine the FAST TCP protocol and the ECN protocol to overcome the inability to play live video smoothly.

Contents of the invention

The purpose of the present invention is to provide a method and system for optimizing the transmission performance of a live video network based on software definition. The present invention makes full use of the software-defined method to control the characteristics of the entire network control plane, calculates the congestion price vector of the entire network, and passes the congestion notification protocol The notification sending network information sending port; the network information sending port actively adjusts the sending rate according to the congestion price vector to avoid network congestion, find and solve the optimization problem of high-speed network transmission performance from the internal cause, and realize the smooth playback of live video.

To achieve the above object, the present invention provides the following scheme:

The present invention provides a method for optimizing the transmission performance of a live video network based on software definition. The optimization method includes:

Obtain global network link operation information;

Determine sample data according to the global network link operation information;

According to the sample data, calculate the guiding price of network link congestion;

Obtain the current network link congestion status;

determining a current network link congestion price vector according to the network link congestion guidance price and the current network link congestion state;

Send the current network link congestion price vector to the network information sending port, and adjust the sending rate of the network information sending port according to the current network link congestion price vector received by the network information sending port.

Optionally, the acquisition of global network link operation information specifically includes:

Send a detection packet to the switch device, and receive a return detection packet sent by the switch device according to the sent detection packet;

According to the time interval between sending the detection packet and receiving the return detection packet, judge the current link congestion state; if the time interval is greater than the set threshold, then judge that the original link is congested or unavailable; otherwise, judge the original link Some links are in normal state;

According to the current link congestion state, global network link operation information is acquired.

Optionally, the determining sample data specifically includes:

Calculate the time interval between sending a probe packet and receiving a return probe packet;

According to the time interval, it is judged whether the time interval is greater than the set threshold; if not, the normal data of the link is obtained;

Determine sample data according to the link normal data.

Optionally, the calculating the network link congestion guidance price specifically includes:

Set the mapping relationship between the global network congestion status and the calculation of congestion price parameters;

According to the sample data and the mapping relationship, a neural network model is constructed, wherein the neural network model represents a three-layer neural network whose input layer number is N, the number of hidden layer neurons is M, and the number of output layers is 1. network;

Optimizing the weights and thresholds of each layer connection in the neural network model;

According to the optimized weight and threshold, the network link congestion guidance price is calculated.

Optionally, the adjusting the sending rate of the network information sending port specifically includes:

According to the current network link congestion price vector, set the IP packet of the congestion mark ratio;

Send the IP packet to the network information sending port through the congestion notification protocol;

Extracting the current network link congestion price vector in the IP packet in the network information sending port;

Construct the FAST TCP header option and the IP header option of the sending end;

Obtain the current network operation mode of the network information sending port according to the FAST TCP header option and the IP header option of the sending end;

Determine the size of the current sending window according to the current network operation mode;

Adjust the sending rate of the network information sending port according to the size of the current sending window.

The present invention also provides a software-defined live video network transmission performance optimization system. The optimization system includes:

The link operation information acquisition module is used to obtain the global network link operation information;

A sample data determining module, configured to determine sample data according to the global network link operation information;

A network link congestion guidance price calculation module, configured to calculate a network link congestion guidance price based on the sample data;

The current network link congestion state acquisition module is used to obtain the current network link congestion state;

A current network link congestion price vector determination module, configured to determine a current network link congestion price vector according to the network link congestion guidance price and the current network link congestion state;

A sending rate adjustment module, configured to send the current network link congestion price vector to the network information sending port, and adjust the sending rate of the network information sending port according to the current network link congestion price vector received by the network information sending port .

Optionally, the acquisition of global network link operation information specifically includes:

The data transmission sub-module is used to send the detection packet to the switch device, and receive the return detection packet sent by the switch device according to the sent detection packet;

The current link congestion state judging submodule is used to judge the state of the current link according to the time interval between sending the detection packet and receiving the return detection packet; if the time interval is greater than the set threshold, it is judged that the original link is Congested state or unavailable state; otherwise, it is judged that the original link is in normal state;

The global network link operation information acquisition sub-module is configured to acquire global network link operation information according to the state of the current link.

Optionally, the determining sample data specifically includes:

The time interval calculation sub-module is used to calculate the time interval between sending the detection packet and receiving the return detection packet;

The time interval judging submodule is used to judge whether the time interval is greater than the set threshold according to the time interval; if not, obtain link normal data;

The sample data determining submodule is configured to determine sample data according to the link normal data.

Optionally, the calculating the network link congestion guidance price specifically includes:

The relationship setting sub-module is used to set the mapping relationship between the global network congestion state and the calculation congestion price parameter;

The neural network construction submodule is used to construct a neural network model according to the sample data and the mapping relationship, wherein the neural network model represents that the number of input layers is N, the number of neurons in the hidden layer is M, and the output layer A three-layer neural network whose number is 1;

Weight and threshold optimization sub-module, used to optimize the weight and threshold of each layer connection in the neural network model;

The network link congestion guidance price calculation sub-module is used to calculate the network link congestion guidance price according to the optimized weight and threshold.

Optionally, the adjusting the sending rate of the network information sending port specifically includes:

The IP packet setting submodule is used to set the IP packet of the congestion mark ratio according to the current network link congestion price vector;

The IP packet sending submodule is used to send the IP packet to the network information sending port through the congestion notification protocol;

An extracting submodule, configured to extract the current network link congestion price vector in the IP packet in the network information sending port;

FAST TCP header option and IP header option construction submodule, used to construct the TCP header option and IP header option of the sending end;

The current network operation mode analysis submodule is used to obtain the current network operation mode of the network information sending port according to the FAST TCP header option and the IP header option of the sending end;

The current sending window size determining submodule is used to determine the current sending window size according to the current network operation mode;

The sending rate adjusting sub-module is configured to adjust the sending rate of the network information sending port according to the size of the current sending window.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the present invention determines the sample data by acquiring the global network link operation information, and calculates the network link congestion guidance price according to the determined sample data; obtains the current network link congestion state, and calculate the current network link congestion price vector according to the network link congestion guidance price and the current network link congestion state; send the current network link congestion price vector to the network information sending port, and According to the current network link congestion price vector received by the network information sending port, adjust the sending rate of the network information sending port to avoid network congestion, find out from the internal cause to solve the problem of high-speed network transmission performance optimization, and improve the data transmission of the live video network performance.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings in the following description are only some of the present invention. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without paying creative labor.

FIG. 1 is a flowchart of a software-defined live video network transmission performance optimization method in an embodiment of the present invention;

2 is a structural diagram of a software-defined live video network transmission performance optimization system in an embodiment of the present invention;

Fig. 3 is the neural network model structural diagram in the embodiment of the present invention;

Fig. 4 is a network operation mode diagram in the embodiment of the present invention.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Embodiment one

The present invention provides a method for optimizing the transmission performance of a live video network based on software definition, as shown in FIG. 1 , specifically including:

Step 101: Obtain global network link operation information, including:

The controller periodically sends Link Layer Discovery Protocol (LLDP) to all switching devices to collect connection information of network switching devices, and build a global network topology map based on the connection information between the controller and network switching devices;

Sending a detection packet to a switch device according to the global network topology diagram, and receiving a return detection packet sent by the switch device according to the sent detection packet;

Use the time interval between sending the detection packet and receiving the returned detection packet to judge the current link network congestion state, which specifically includes: recording the time when the detection packet is sent and starting the timer. If the time interval is greater than the set threshold, then judge the cause Some links are congested or unavailable; otherwise, the original link is judged to be in a normal state;

According to the obtained current link network congestion status, the global network link operation information is obtained.

Step 102: Determine sample data, including:

Calculate the time interval between sending a probe packet and receiving a return probe packet;

According to the time interval, it is judged whether the time interval is greater than the set threshold; if not, the normal data of the link is obtained;

Determine sample data according to the link normal data.

Step 103: Calculate the guide price of network link congestion, including:

Set the mapping relationship between the global network congestion state and the calculation of congestion price parameters, specifically including: according to the openness and programmability of the software-defined network architecture, the network administrator sets the mapping relationship between the global network congestion state and the calculation of congestion price parameter vectors based on experience ; When link resources are not tight, select a lower congestion price calculation parameter vector; when each link resource is tight, choose a significantly higher congestion price calculation parameter vector, as shown in Table 1.

Table 1: Relationship between link load and congestion guidance price

According to the sample data and the mapping relationship, a neural network model is constructed, wherein the neural network model represents a three-layer BP whose input layer number is N, the number of hidden layer neurons is M, and the output layer number is 1. The neural network specifically includes: on the basis of the OpenFlow1.3 protocol, using the OpenFlow channel of the controller and the switch, using the controller and the switch to exchange three kinds of encapsulated controller-to-switch, asynchronous and symmetric packets, constructing the OpenFlow switch and control The data structure of the device interaction phase information. The above-mentioned data structure includes the switches periodically uploading link load information s ₁ , s ₂ , s...,s _n of n switches in the data center network, actual application effects, and the controller macroscopically guiding the exchange congestion machine price setting strategy. The input of the sample data is the link load information of each switch, and the output is the defuzzified congestion price guidance strategy adopted, and according to the defuzzification congestion price guidance strategy, an input layer as shown in the figure is constructed in the central controller The three-layer BP neural network whose number is n (the number of switches), the number of hidden layer neurons is m, and the number of output layers is 1, as shown in Figure 3, and the connection weights of each layer of the BP neural network are set And the threshold is first randomly initialized to a value between [0, 1].

According to the optimized weights and thresholds, the optimized neural network model is obtained, which specifically includes: as more and more original data, in order to improve the convergence speed of the neural network, the fuzzy C-average clustering algorithm will be used to correlate Attributes or properties are processed into the same classification to generate multiple groups and reduce duplicate data.

The idea of fuzzy C-average grouping method is as follows:

The objective function of Fuzzy C-means is as formula (1):

Among them, u _ij is the belonging degree of data point s _i in cluster j, m is any real number whose weight coefficient can be greater than 1, ||s _i -c _j || ² represents the data point s _i and the cluster center c The distance function of _j is generally calculated using Euclidean distance. Fuzzy C-means minimizes the iterative membership function and cluster center to obtain better clustering results, and updates the membership function u _ij according to formula (2):

Update the cluster center function c _j according to formula (3):

like Then the iteration stops. Here ε is the preset error tolerance value, and k is the number of iterations.

Specific steps are as follows:

1. Initial setting of parameters: set the sample data data, the number of clusters c, the number of iterations k, and initialize the set matrix U of belongingness;

2. Update the cluster center: calculate the cluster center c _j through the formula (3).

3. Calculating group membership: update the new group membership degree matrix through formula (2).

4. Termination condition: when , the iteration stops, otherwise go back to step 2 and repeat.

According to the optimized weights and thresholds, the congestion price calculation strategy is obtained, which specifically includes: offline training of each connection weights and thresholds of the BP neural network according to the sample data to obtain an optimal solution.

According to the optimal solution, the actual load submitted by each switch and the neural network model obtained from the above training, the current state of the switch is calculated, and the congestion price calculation strategy to be adopted by the current switch is deduced according to the fuzzy rules.

According to the congestion price calculation strategy, calculating the network link congestion guidance price specifically includes: at the switch level, according to the macro guidance policy for calculating the congestion price at the upper layer, researching the relationship between the parameters in the congestion price calculation formula and the controller's macro guidance strategy, Calculate the guide price of network link congestion;

Step 104: Obtain the current network link congestion state;

Step 105: Determine the current network link congestion price vector, specifically including: according to the network link congestion guidance price and the current network link congestion state, research the relationship between the parameters in the congestion price calculation formula and the controller's macro guidance strategy , combined with the link congestion state of the high-speed network given by the controller, select the congestion price parameter vector B, calculate the parameter price of each switch according to p=f(B, Q), and determine the current network link congestion price vector, Q is the link The current congestion state of the road, where p=f(B,Q), where B is a constant vector, and Q is the time difference vector between sending a probe packet and receiving a returned packet

Step 106: Adjust the sending rate of the network information sending port, specifically including:

According to the current network link congestion price vector, the IP packet of the congestion mark ratio is set, which specifically includes: at the switch level, using the programmable characteristics of the software-defined architecture to set the IP packet of the congestion mark ratio;

Add an explicit congestion notification algorithm, so that the switch can accurately feed back the IP packet of the congestion mark ratio to the high-speed network transmission control protocol at the sending end, and then send it to the network information sending port;

Extracting the current network link congestion price vector in the IP packet in the network information sending port;

Constructing the TCP header option and the IP header option of the sending end specifically includes: adopting the FAST TCP protocol at the network information sending port, and adding the TCP header option and the IP header option at the FAST packet header and the IP packet header. The content of the TCP header option and the IP header option includes the interaction information between the FAST sender and the link-end routing device, which can specifically include the following content: current protocol parameter values, historical protocol parameter values, current network bottleneck link congestion prices, and historical network Bottleneck link congestion price; among them, in order to ensure fairness, each sender endows the same purchasing power.

Obtain the current network operation mode of the network information sending port, specifically including: at the network information sending port, statistically obtain the sending window size, round-trip delay, network congestion price and the change value of the above data, perform feature analysis, extraction and feature composite design, and coordinate Analyze and obtain the network operation mode as shown in Figure 4;

According to the current network operation mode, determine the size of the current sending window, specifically including: determining the changing size of the sending window according to different modes

Adjusting the sending rate of the network information sending port according to the size of the current sending window specifically includes: adjusting the sending rate of the sending end according to the size of the current sending window. As shown in Figure 4, when the error is large (e>e ₁ ), corresponding to area ①, it is necessary to take a larger window change step size; when the error is small ((e<e ₂ ), and ), corresponding to area ⑤, then take a smaller window change step.

Through the above-mentioned embodiments, the present invention realizes avoiding network congestion, finds and solves the optimization problem of high-speed network transmission performance from the internal cause, and improves the data transmission performance of the live video network.

The present invention also provides a software-defined live video network transmission performance optimization system, as shown in Figure 2, specifically including: link operation information acquisition module 201, sample data determination module 202, network link congestion guidance price calculation module 203, The current network link congestion status acquisition module 204 , the current network link congestion price vector determination module 205 and the sending rate adjustment module 206 .

The link operation information acquisition module 201 is used to acquire global network link operation information, specifically including:

The data transmission sub-module is used to send the detection packet to the switch device, and receive the return detection packet sent by the switch device according to the sent detection packet;

The current link congestion state judging submodule is used to judge the state of the current link according to the time interval between sending the detection packet and receiving the return detection packet; if the time interval is greater than the set threshold, it is judged that the original link is Congested state or unavailable state; otherwise, it is judged that the original link is in normal state;

The global network link operation information acquisition sub-module is configured to acquire global network link operation information according to the state of the current link.

The sample data determining module 202 is configured to determine sample data according to the global network link operation information, specifically including:

The time interval calculation sub-module is used to calculate the time interval between sending the detection packet and receiving the return detection packet;

The time interval judging submodule is used to judge whether the time interval is greater than the set threshold according to the time interval; if not, obtain link normal data;

The sample data determining submodule is configured to determine sample data according to the link normal data.

The network link congestion guidance price calculation module 203 is configured to calculate the network link congestion guidance price according to the sample data, specifically including:

The relationship setting sub-module is used to set the mapping relationship between the global network congestion state and the calculation congestion price parameter;

The neural network construction submodule is used to construct a neural network model according to the sample data and the mapping relationship, wherein the neural network model represents that the number of input layers is N, the number of neurons in the hidden layer is M, and the output layer A three-layer neural network whose number is 1;

Weight and threshold optimization sub-module, used to optimize the weight and threshold of each layer connection in the neural network model;

The network link congestion guidance price calculation sub-module is used to calculate the network link congestion guidance price according to the optimized weight and threshold.

The current network link congestion status acquisition module 204 is used to acquire the current network link congestion status;

The current network link congestion price vector determining module 205 is configured to determine the current network link congestion price vector according to the network link congestion guidance price and the current network link congestion state;

The sending rate adjustment module 206 is configured to send the current network link congestion price vector to the network information sending port, and adjust the sending of the network information sending port according to the current network link congestion price vector received by the network information sending port rate, including:

The IP packet setting submodule is used to set the IP packet of the congestion mark ratio according to the current network link congestion price vector;

The IP packet sending submodule is used to send the IP packet to the network information sending port through the congestion notification protocol;

An extracting submodule, configured to extract the current network link congestion price vector in the IP packet in the network information sending port;

FAST TCP header option and IP header option construction submodule, used to construct the TCP header option and IP header option of the sending end;

The current network operation mode analysis submodule is used to obtain the current network operation mode of the network information sending port according to the FAST TCP header option and the IP header option of the sending end;

The current sending window size determining submodule is used to determine the current sending window size according to the current network operation mode;

The sending rate adjusting sub-module is configured to adjust the sending rate of the network information sending port according to the size of the current sending window.

Through the above-mentioned embodiments, the present invention realizes avoiding network congestion, finds and solves the optimization problem of high-speed network transmission performance from the internal cause, and improves the data transmission performance of the live video network.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other. As for the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and for relevant details, please refer to the description of the method part.

In this paper, specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; meanwhile, for those of ordinary skill in the art, according to the present invention Thoughts, there will be changes in specific implementation methods and application ranges. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims (10)

1. a kind of net cast network transmission performance optimization method based on software definition, it is characterised in that the optimization method, Including：

Obtain global network link operation information；

According to the global network link operation information, sample data is determined；

According to the sample data, calculating network link congestion guide price；

Obtain current network link congestion state；

According to the network link congestion guide price and the current network link congestion state, determine that current network link is gathered around Plug price vector；

The current network link congestion price vector is sent to network information sending port, and according to network information transmitting terminal The current network link congestion price vector that mouth is received, adjusts the transmission rate of the network information sending port.

2. optimization method according to claim 1, it is characterised in that the acquisition global network link operation information, tool Body includes:

Detection bag is sent to switch device, and receive the return detection that the switch device is sent according to transmission detection bag Bag；

According to sending detection bag and the time interval for returning detection bag being received, current ink congestion state is judged；If when described Between be spaced more than given threshold, then judge original link for congestion state or down state；Otherwise then judge original chain Road is normal condition；

According to the current ink congestion state, global network link operation information is obtained.

3. optimization method according to claim 1, it is characterised in that the determination sample data, specifically includes：

Calculate the time interval for sending detection bag and receiving return detection bag；

According to the time interval, judge time interval whether more than given threshold；If it is not, then obtaining link normal data；

According to the link normal data, sample data is determined.

4. optimization method according to claim 1, it is characterised in that the calculating network link congestion guide price, tool Body includes：

Global network congestion state is set and calculates the mapping relations of congestion price parameter；

According to the sample data and the mapping relations, neural network model is built, wherein described neural network model represents One input layer number is N, and hidden neuron number is M, and output layer number is 1 three-layer neural network；

Optimize weights and threshold value that each layer in the neural network model connects；

According to the weights after optimization and threshold value, calculating network link congestion guide price.

5. optimization method according to claim 1, it is characterised in that the sending out of the adjustment network information sending port Transmission rate, specifically includes：

According to the current network link congestion price vector, the IP bag of congestion flag ratio is set；

By congestion notification agreement, the IP bag will be set and be sent to network information sending port；

Extract the current network link congestion price vector in IP bag in the network information sending port；

Build FAST TCP header options and the IP header options of the transmitting terminal；

According to FAST TCP header options and the IP header options of the transmitting terminal, network information sending port current network is obtained Operational mode；

According to the current network operational mode, currently transmitted window size is determined；

According to currently transmitted window size, the transmission rate of the network information sending port is adjusted.

6. a kind of net cast network transmission performance based on software definition optimizes system, it is characterised in that the optimization system, Including：

Link operation information acquisition module, for obtaining global network link operation information；

Sample data determining module, for according to the global network link operation information, determining sample data；

Network link congestion guide price computing module, for according to the sample data, calculating network link congestion guiding price Lattice；

Current network link congestion state acquisition module, for obtaining current network link congestion state；

Current network link congestion price vector determining module, for according to the network link congestion guide price and described work as Front network link congestion state, determines current network link congestion price vector；

Transmission rate adjusting module, for being sent to network information transmitting terminal by the current network link congestion price vector Mouthful, and the current network link congestion price vector received according to network information sending port, adjust the network information and send The transmission rate of port.

7. the optimization system according to claim 6, it is characterised in that the acquisition global network link operation information, tool Body includes:

Data transfer submodule, for sending detection bag to switch device, and receives the switch device according to transmission spy Survey the return detection bag that bag sends；

Current ink congestion state judging submodule, returns between the time that detection is wrapped for wrapping and receiving according to transmission detection Every judging the state of current ink；If the time interval is more than given threshold, judge original link for congestion state or Down state；Otherwise then judge original link for normal condition；

Global network link operation information acquisition submodule, for the state according to the current ink, obtains global network chain Road operation information.

8. optimization system according to claim 6, it is characterised in that the determination sample data, specifically includes：

Time interval calculating sub module, for calculating the time interval for sending detection bag and receiving return detection bag；

Whether time interval judging submodule, for according to the time interval, judging time interval more than given threshold；If No, then obtain link normal data；

Sample data determination sub-module, for according to the link normal data, determining sample data.

9. optimization system according to claim 6, it is characterised in that the calculating network link congestion guide price, tool Body includes：

Relation arranges submodule, for arranging global network congestion state and calculating the mapping relations of congestion price parameter；

Neutral net builds submodule, for according to the sample data and the mapping relations, building neural network model, its Described in neural network model represent an input layer number for N, hidden neuron number is M, and output layer number is three layers of 1 Neutral net；

Weights and threshold optimization submodule, for optimizing weights and the threshold value of the connection of each layer in the neural network model；

Network link congestion guide price calculating sub module, for according to the weights after optimization and threshold value, calculating network chain Road congestion guide price.

10. optimization system according to claim 6, it is characterised in that the adjustment network information sending port Transmission rate, specifically includes：

IP bag arranges submodule, for according to the current network link congestion price vector, arranging the IP of congestion flag ratio Bag；

IP bag sending submodule, for passing through congestion notification agreement, will set the IP bag and be sent to network information sending port；

Extracting sub-module, for extract the current network link congestion price in the network information sending port in IP bag to Amount；

FAST TCP header options and IP header options build submodule, for build the transmitting terminal TCP header options and IP header options；

Current network Operation Mode Analysis submodule, for selecting according to the FAST TCP header options of the transmitting terminal and IP stem , obtain network information sending port current network operational mode；

Currently transmitted window size determination sub-module, for according to the current network operational mode, determining currently transmitted window Size；

Transmission rate adjusts submodule, for according to currently transmitted window size, adjusting sending out for the network information sending port Transmission rate.