CN103414650A - Routing method and device for congestion avoidance - Google Patents

Abstract

Provided is a routing method and device for congestion avoidance. The method comprises the steps of receiving LSA messages sent by all first network elements in a network, structuring a global network topological graph, receiving NCA messages sent by all the first network elements in the network, obtaining global congestion information, obtaining the optimal path of other nodes in a network accessed by each node, obtaining the maximum inflow speed allowed by the optimal path, removing all nodes in congestion states in the network from the topological graph to obtain a network topological graph of a non-congestion state according to the global congestion information, obtaining the optimal path in non-congestion according to the network topological graph of the non-congestion state, obtaining a forwarding strategy according to the obtained optimal path, the maximum inflow speed of the optimal path and the optimal path in the non-congestion, sending the forwarding strategy to the corresponding first network elements to enable the first network elements to obtain flow chart entries according to the forwarding strategy, and carrying out storage. Therefore, congestion is avoided, and data transmission efficiency is improved.

Description

A kind of method for routing of avoid congestion and device

Technical field

The present invention relates to communication technical field, particularly a kind of method for routing of avoid congestion and device.

Background technology

Along with the growth of internet scale, the user on the Internet and application all increase fast, have congestedly become a very important problem.In recent years, carried out a large amount of research work in the congestion control field, congestion control algorithm can be divided into two major parts, that is: at the source algorithm used on end system and the link algorithm used on the network equipment.

The source algorithm can be understood as, and is occurring when congested, in source, to reduce the speed that sends data, thereby makes a large amount of connections can shared link.By source, the perception of network current state and detection are responded to the window size of adjusting the generation grouping, unquestionable, the extensibility that this algorithm provides has promoted the fast development that the Internet is early stage.Along with the development of technology, the cognition of the effect that the node in network is brought into play deeply, has proposed Queue Management Algorithm gradually thus.Queue Management Algorithm is, by management routing queue length, each formation is arranged to a maximum, then receives bag and enters formation until queue length reaches maximum, and the packet of follow-up arrival will be dropped, and namely abandon tail algorithm (DT, Drop Tail).

In prior art, also proposed Active Queue for Management Mechanism (AQM, Active Queue Management), wherein, early stage random drop (RED, Random Early Detection) is based on most representative scheme in the AQM of queue length.The RED algorithm comes Sampling network congested by the average queue length q that detects intermediate node, and it is congested by the probability dropping with certain, to divide into groups to notify the TCP transmit leg to occur.Packet dropping probability in the RED algorithm is a function about q, and it below certain level, can avoid burst type to abandon continuously the packet from same connection by the packet leashes that will receive.Shown in following formula (1):

$P=\left\{\begin{array}{cc}0,& q\&le;q_{\min }\\ P_{\max }\frac{q-q_{\min }}{q_{\max }-q_{\min }},& q_{\min }<q<q_{\max }\\ 1,& q_{\max }<q\end{array}\right.---\left(1\right)$

Wherein, q _min, q _max, be minimum threshold and the max-thresholds of formation, by formula (1), just can calculate the probability that grouping abandons.As shown in Figure 1, the drop probability of RED algorithm meets piecewise linear relationship with the variation of average queue length.

Technical scheme from above-mentioned existing solution congestion problems, abandon the method for this simple avoid congestion of tail algorithm, the phenomenon of synchronous packet loss can occur in network, each data flow is subject to the impact of DT algorithm, constantly increase and reduce window, thereby cause waving of network cycle between congested and idle two states, cause the decline of link utilization.The AQM algorithm can be avoided in network synchronous packet loss phenomenon, but this algorithm, at network, the waste that packet loss can cause Internet resources does not occur when congested just to start in advance.

Therefore, it is a kind of more reasonable that prior art can not provide, the technical scheme of more optimizing.

Summary of the invention

The embodiment of the present invention provides a kind of method for routing and device of avoid congestion, takes full advantage of the formation buffer memory of node, simultaneously again can avoid congestion packet loss phenomenon, from source, solve the problem of network congestion.

First aspect, provide a kind of method for routing of avoid congestion, comprising:

The first network element obtains information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained, and the LSA message is sent to controller;

Described the first network element obtains congestion state bulletin NCA message, and described NCA message is sent to described controller, wherein, in described NCA message, includes the congestion information of described the first network element.

In the possible implementation of the first, according to first aspect, the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,i}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state.

In the possible implementation of the second, in conjunction with first aspect or the possible implementation of the first, described method also comprises:

Described the first network element sends the numerical value of the streaming rate that enters and leave described the first network element, and the described numerical value that enters and leave the streaming rate of described the first network element obtains structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

Second aspect, provide a kind of method for routing of avoid congestion, comprising:

Receive the LSA message that in network, each first network element sends; According to described LSA message, construct overall network topology figure;

Receive the NCA message that in network, each first network element sends; According to described NCA message, obtain global congestion information;

According to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

According to described global congestion information, and the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows obtained;

According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, according to described forwarding strategy, obtains stream table list item, and send to corresponding the first network element;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the described optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested.

In the possible implementation of the first, according to second aspect, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, also specifically comprises:

The speed v of the data flow received in source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

In the possible implementation of the second, according to first aspect, the perhaps possible implementation of the first of first aspect, each NCA message comprises the congestion information of first network element, the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,i}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

According to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

In the third possible implementation, in conjunction with the possible implementation of the second, described according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely

Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

In the 4th kind of possible implementation, in conjunction with second aspect, in second aspect first, possible implementation is to possible implementation in second aspect the 3rd, any one, the optimal path that described basis is obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, be specially:

According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in the first network element is tcp data stream.

In the 5th kind of possible implementation, in conjunction with second aspect, described method also comprises:

Receive the streaming rate that enters and leave each first network element that each first network element sends;

Described according to described overall network topology figure, to obtain after each node goes to the optimal path of other node in network, described method also comprises:

When the data flow received in the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

According to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in the first network element is UDP message stream;

Forwarding strategy when described data flow when receiving in the first network element is UDP message stream, be made as stream table list item, sends to the first network element of described correspondence.

In the 6th kind of possible implementation, in conjunction with the 5th kind of possible implementation of second aspect, when described data flow when receiving in the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length and is less than, maximumly allow queue length, and the ingress rate of the first network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

The third aspect, provide a kind of method for routing of avoid congestion, comprising:

Receiving data stream; Stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception,

Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item;

Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

In the possible implementation of the first, according to the third aspect, described another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward, before, described method also comprises:

Speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

In the possible implementation of the second, according to the third aspect, described method also comprises:

If the described optimal path recorded in described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

In the third possible implementation, in conjunction with the third aspect or possible implementation or the possible implementation of the second of the first, described stream table list item according to storage, before the data flow of described reception was divided into to two data flow forwardings, described method also comprised:

Whether the data flow that judges described reception is tcp data stream, if tcp data stream is carried out described stream table list item according to storage, the data flow of described reception is divided into to two data flow forwardings.

In the 4th kind of possible implementation, in conjunction with the third possible implementation of the third aspect, described method also comprises:

When the data flow of the described reception of judgement is not tcp data stream, according to record in described stream table list item, carry out and forward, specifically comprise:

The queue length of packet is more than or equal to threshold value length in described the first network element, and is less than and maximum allows queue length, and the ingress rate of described the first network element is greater than while leaving speed, with ρ _iThe probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iDrop probabilities for record the first network element i in described stream table list item;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

Fourth aspect, provide a kind of network element, and described network element comprises: the first acquiring unit, the first transmitting element, second acquisition unit, and the second transmitting element;

Described the first acquiring unit, be used to obtaining information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained;

Described the first transmitting element, for sending to controller by the LSA message;

Described second acquisition unit, be used to obtaining congestion state bulletin NCA message, wherein, include the congestion information of described network element in described NCA message;

Described the second transmitting element, for sending to described controller by described NCA message.

In the possible implementation of the first, according to fourth aspect, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

In the possible implementation of the second, in conjunction with fourth aspect, the perhaps possible implementation of the first, described the second transmitting element, also for transmission, enter and leave the numerical value of the streaming rate of described network element, the described numerical value that enters and leave the streaming rate of described network element, obtain structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

The 5th aspect, provide a kind of controller, and described controller comprises:

The first receiving element, the first structural unit, the second receiving element, the 3rd acquiring unit, the 4th acquiring unit, the 5th acquiring unit, the 6th acquiring unit, the 7th acquiring unit, the 3rd transmitting element;

Described the first receiving element, the LSA message sent for receiving each first network element of network;

Described the first structural unit, for according to described LSA message, construct overall network topology figure;

Described the second receiving element, the NCA message sent for receiving each first network element of network;

Described the 3rd acquiring unit, for according to described NCA message, obtain global congestion information;

Described the 4th acquiring unit, for according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

Described the 5th acquiring unit, for according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows;

Described the 6th acquiring unit, for according to described global congestion information, by all nodes of congestion state that are in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

Described the 7th acquiring unit, for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, obtains stream table list item according to described forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested;

Described the 3rd transmitting element, send to described the first network element for the described stream table list item that will obtain.

In the possible implementation of the first, according to the 5th aspect, in described the 7th acquiring unit for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, also specifically comprise:

The speed v of the data flow received when source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

In the possible implementation of the second, in conjunction with the 5th aspect, each the NCA message received in the second receiving element comprises the congestion information of first network element, and the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

In the 3rd acquiring unit, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

In the third possible implementation, in conjunction with the possible implementation of the second, in described the 5th acquiring unit, be used for according to described global congestion information, with the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely

Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

In the 4th kind of possible implementation, in conjunction with the 5th aspect or the possible implementation of the first to any one of the third possible implementation, in described the 7th acquiring unit for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, be specially:

Described according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding the first network element is tcp data stream.

In the 5th kind of possible implementation, in conjunction with the 5th aspect, described the second receiving element, the streaming rate that enters and leave each first network element also sent be used to receiving each first network element;

Described controller also comprises:

The second structural unit and the 4th transmitting element;

Described the second structural unit, be used at described the 4th acquiring unit according to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

Described the 4th transmitting element, for by the described data flow received when corresponding the first network element, being the UDP message forwarding strategy in when stream, be made as stream table list item, sends to the first network element of described correspondence.

In the 6th kind of possible implementation, in conjunction with the possible implementation in the 5th kind of the 5th aspect, while in described the second structural unit, being UDP message stream for the data flow received when corresponding the first network element, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and the first network element that allows and be greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

The 6th aspect, provide a kind of network element, and described network element comprises: the 4th receiving element, and the first retransmission unit;

Described the 4th receiving element, for receiving data stream;

Described the first retransmission unit, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

In the possible implementation of the first, according to the 6th aspect, described network element also comprises: indicate unit,

Described sign unit, for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

In the possible implementation of the second, according to the 6th aspect, described network element also comprises: the second retransmission unit,

Described the second retransmission unit, if the described optimal path recorded for described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

In the third possible implementation, in conjunction with the 6th aspect or the possible implementation of the first, or the possible implementation of the second, described network element also comprises: the second judging unit,

Described the second judging unit, at the stream table list item of described the first retransmission unit according to storage, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described the first retransmission unit, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

In the 4th kind of possible implementation, in conjunction with possible implementation in the 6th aspect the 3rd, described network element also comprises: the 3rd retransmission unit,

Described the 3rd retransmission unit, while for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iFor in described stream table list item, recording the drop probabilities of network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

The 7th aspect, provide a kind of network element, and described network element comprises the input unit be connected respectively on bus, output device, memory and processor;

Described processor, for the information of neighbor nodes according to obtaining from input unit, according to the information of neighbor nodes link generation state bulletin LSA message obtained; From input unit, obtain congestion state bulletin NCA message, wherein, in described NCA message, include the congestion information of described network element;

Described output device, for sending to controller by the LSA message; Described NCA message is sent to described controller.

In the possible implementation of the first, according to the 7th aspect, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

In the possible implementation of the second, according to the 7th aspect or the possible implementation of the 7th aspect the first, described output device, also for transmission, enter and leave the numerical value of the streaming rate of described network element, the described numerical value that enters and leave the streaming rate of described network element, obtain structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

Eight aspect, provide a kind of controller, and described controller comprises the input unit be connected respectively on bus, output device, memory and processor;

Described input unit, the LSA message sent for receiving each first network element of network; Receive the NCA message that in network, each first network element sends;

Described processor, for according to described LSA message, construct overall network topology figure; According to described NCA message, obtain global congestion information; According to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, obtains stream table list item according to described forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested;

Described output device, send to described the first network element for the described stream table list item that will obtain.

In the possible implementation of the first, according to eight aspect, in described processor for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, also specifically comprise:

The speed v of the data flow received when source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

In the possible implementation of the second, in conjunction with the possible implementation of the first, each the NCA message received in described input unit comprises the congestion information of first network element, and the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

In described processor, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

In the third possible implementation, in conjunction with the possible implementation of eight aspect the second, in described processor, be used for according to described global congestion information, with the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

In the 4th kind of possible implementation, in conjunction with eight aspect or the possible implementation of the first to the third possible implementation any one, in described processor for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, be specially:

Described according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding the first network element is tcp data stream.

In the 5th kind of possible implementation, in conjunction with the 4th kind of possible implementation of eight aspect, described input unit, the streaming rate that enters and leave each first network element also sent be used to receiving each first network element;

Described processor also for:

According to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

Forwarding strategy when described data flow when receiving in corresponding the first network element is UDP message stream, be made as stream table list item, sends to the first network element of described correspondence.

In the 6th kind of possible implementation, in conjunction with the 5th kind of possible implementation of eight aspect, while in described processor, being UDP message stream for the data flow received when corresponding the first network element, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and the first network element that allows and be greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

The 9th aspect, provide a kind of network element, and described network element comprises the input unit be connected respectively on bus, output device, memory and processor;

Described input unit, for receiving data stream;

Described processor, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

In the possible implementation of the first, according to the 9th aspect, described processor, also for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

In the possible implementation of the second, in conjunction with the possible implementation of the 9th aspect the first, described processor, if the described optimal path also recorded for described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

In the third possible implementation, in conjunction with the 9th aspect or possible implementation or the possible implementation of the second of the first, it is characterized in that, described processor, also for described the first retransmission unit according to the storage stream table list item, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described processor, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

In the 4th kind of possible implementation, in conjunction with the third possible implementation, described processor, while also for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iFor in described stream table list item, recording the drop probabilities of network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

The embodiment of the present invention provides a kind of method for routing and device of avoid congestion, this scheme middle controller is according to from each switch network, obtaining LSA message and NCA message, construct overall network topology figure and global congestion information, according to overall network topology figure and global congestion information, construct switch at two paths that are under congestion state, be about to congested data flow and be divided into two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

The accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, below will the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain according to these accompanying drawings other accompanying drawing.

Fig. 1 is the drop probability piecewise linear relationship schematic diagram of RED algorithm in prior art;

Fig. 2 is message repeating simplified schematic diagram between Open Flow switch and controller in the SDN network;

Fig. 3 is that the stream of storing in Open Flow switch means the sketch of anticipating;

Fig. 4 is a kind of SDN network simplified schematic diagram;

Fig. 5 is the method for routing general flow chart of a kind of avoid congestion of providing of the embodiment of the present invention;

Fig. 6 is the method for routing general flow chart that the two a kind of avoid congestions that provide are provided the embodiment of the present invention;

Fig. 7 is the method for routing general flow chart that the four a kind of avoid congestions that provide are provided the embodiment of the present invention;

Fig. 8 a is the method for routing flow process part sketch that the five a kind of avoid congestions that provide are provided the embodiment of the present invention;

Fig. 8 b is the method for routing flow process part sketch that the five a kind of avoid congestions that provide are provided the embodiment of the present invention;

Fig. 9 is the LSB message simplified schematic diagram that each switch sends to controller;

Figure 10 is the NCA message simplified schematic diagram that each switch sends to controller;

Figure 11 is that bottom overall network topology figure illustrates sketch;

Figure 12 is the optimal path simplified schematic diagram from the source node to the destination node of the non-congested network of controller structure;

Figure 13 is stream table list item simplified schematic diagram in switch;

Figure 14 is the stream table list item simplified schematic diagram of distinguishing tcp data stream and UDP message stream in switch;

Figure 15 is a kind of network element simplified schematic diagram that the embodiment of the present invention six provides;

Figure 16 is a kind of controller simplified schematic diagram that the embodiment of the present invention seven provides;

Figure 17 is a kind of network element simplified schematic diagram that the embodiment of the present invention nine provides;

Figure 18 is a kind of network element simplified schematic diagram that the embodiment of the present invention ten provides;

Figure 19 is a kind of controller simplified schematic diagram that the embodiment of the present invention 11 provides;

Figure 20 is a kind of network element simplified schematic diagram that the embodiment of the present invention 12 provides.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Based on the embodiment in the present invention, those of ordinary skills, not making under the creative work prerequisite the every other embodiment obtained, belong to the scope of protection of the invention.

Before the technical scheme that the embodiment of the present invention is provided is elaborated, software defined network (the SDN applicable to this programme, Software Defined Networking) describe, SDN is a change for conventional network equipment and framework, and its core concept is forwarded and the separating of network control.Open flows (Open Flow) can be regarded a kind of specific implementation of SDN thought as.As shown in Figure 2, the message repeating process that Open Flow will be controlled by switch and router fully, be converted into by Open Flow switch (Open Flow Switch) and controller (Controller) and jointly complete, thereby realized data retransmission and separating that route is controlled.Controller can be controlled the stream table (stream means the meaning sketch as shown in Figure 3) in Open Flow switch by the interface operation provided in advance, thereby reaches the purpose of controlling data retransmission.

As shown in Figure 4, a kind of SDN network, comprise a controller and a plurality of switch.Each switch is connected with controller by escape way (as dotted arrow in Fig. 2).Switch is responsible for inquiring about stream table forwarding data, and controller is responsible for monitor network and is constructed the stream table.When data flow is flowed through as shown in Figure 2 Local Area Network, at each switch place, stream table in the inquiry switch, if this data flow was not registered, switch sends to controller by this traffic flow information (being mainly the data flow header information), by controller, generate forwarding strategy for this stream, and make stream table list item according to forwarding strategy; If the coupling of the stream table in this data flow and switch, switch carries out associative operation according to the action of stream table by this data flow.It will be appreciated that, Local Area Network under the SDN network architecture shown in Figure 4 is not the restriction to the embodiment of the present invention, be only a kind of explanation of being convenient to understand, the technical scheme that the embodiment of the present invention provides can also indiscriminately be generalized in various types of SDN networks.

Below in conjunction with specific embodiment, the technical scheme that the embodiment of the present invention provides is elaborated.

Embodiment mono-

The embodiment of the present invention provides a kind of method for routing of avoid congestion, the method is the operating process of writing from the angle of each switch network, wherein, it is switch that the said equipment of the present embodiment does not limit, and also can using the first network element as executive agent, and this first network element can be switch, it can be also network node, perhaps a kind of controller etc., should not be construed herein to the embodiment of the present invention now, in literary composition, other embodiment have similar understanding.

As shown in Figure 5, the method comprises:

Step 101, switch obtains information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained; And the LSA message is sent to controller;

Wherein, in network, each switch is regular, or, when neighbor node changes, will pass through the LSA message in the information of neighbor nodes of self, sends to controller, for the information of neighbor nodes of controller according to each switch, constructs overall network topology figure; And, for controller, after the switch by congested (or node) is removed, construct the network topological diagram under non-congestion state.

Step 102, switch are obtained congestion state bulletin NCA message, and described NCA message is sent to described controller, wherein, in described NCA message, include the congestion information of described switch.

It will be appreciated that, the congestion information of the switch that the NCA message that in network, each switch sends comprises, get global congestion information for controller, is convenient to the control that the subsequent control device forwards route.

Wherein, the congestion information of switch is specifically by the congested rate matrix P of described switch _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of switch in zone, p _j,iMean that data flow allows ingress rate from the maximum that switch j enters switch i, wherein, switch i is the switch of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of switch i from switch j; λ _iMean all speed sums that enter switch i, value is

u _iMean all streaming rate sums of leaving switch i, value is

q _iMean the queue length of switch, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of switch length _ThThe time, described switch is in congestion state, as shown in following formula, and when switch is under congestion state, its maximum ingress rate p that allows _j,iReport control, inform controller, this switch is in congestion state.

The method for routing of a kind of avoid congestion by one explanation of above-mentioned the present embodiment, switch in this scheme in network is by the information of neighbor nodes of self and self congestion information, by LSA and NCA, send to controller respectively, by controller to each switch unified management in network, construct overall network topology figure, with global congestion information, make controller according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and described optimal path, the maximum rate of influx that described optimal path allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, and make stream table list item, the stream list item is sent to corresponding switch, make described switch obtain stream table list item, and storage; Wherein, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

The speed v of the data flow received in described corresponding switch is greater than the maximum rate of influx that described optimal path allows, the data flow of described reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as described optimal path, forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with described optimal path forwards according to described non-optimal path when congested.

Therefore, the method of the present embodiment explanation, compared with prior art, avoid simply abandoning the packet over formation max threshold length if abandoned in the tail algorithm, also overcome at network and do not occurred when congested can cause the problem of the waste of Internet resources with regard to starting in advance packet loss, a kind of method of more rational avoid congestion is provided, thereby has guaranteed the quality of transfer of data, for the user provides more high efficiency service.

Preferably, the method can also comprise: the first network element sends the numerical value of the streaming rate that enters and leave described switch, this enters and leaves the numerical value of the streaming rate of described the first network element, obtains structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

Embodiment bis-

The embodiment of the present invention provides a kind of method for routing of avoid congestion, the method of explanation in the method and above-described embodiment one, be based on respectively different executive agents, the controller of take in this embodiment is executive agent, for each switch is planned as a whole forwarding strategy, forwarding strategy is made into to stream table list item, sends to each switch, in return stream table list item in machine.Switch utilizes the forwarding of controller to control, effectively avoid congestion, effective transmission of guaranteed data simultaneously.

The detailed description of the method, shown in accompanying drawing 6, the method comprises:

Step 201, receive the LSA message that in network, each switch sends; According to described LSA message, construct overall network topology figure;

Step 202, receive the NCA message that in network, each switch sends; According to described NCA message, obtain global congestion information;

Wherein, the congestion information that the NCA message that each switch sends comprises, specifically can be with reference to congested rate matrix P in step 102 in above-described embodiment one _iExplanation.At controller, get the congested rate matrix P of each switch _i, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of switch in zone, P _iCongested rate matrix for described switch.

Step 203, according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

Step 204, according to described global congestion information, and described optimal path, obtain the maximum rate of influx that described optimal path allows;

It should be noted that, step 204 can specifically comprise:

From global congestion information and in described optimal path, obtain previous dive k in described optimal path _I-1To down hop k _iMaximum allow ingress rate

Obtain in described optimal path minimum maximum and allow ingress rate, the maximum rate of influx p allowed as described optimal path, namely Wherein, S is the source node of described optimal path, and D is the destination node of described optimal path.

Step 205, according to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

Step 206, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy; Wherein, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in optimal path, having switch to be in congestion state, require the source node switch of this optimal path that the data flow of described reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as described optimal path, forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with described optimal path forwards according to described non-optimal path when congested.

Wherein, need to understand, on path, have switch to be in congestion state, can be understood as this path congestion, therefore, each switch in optimal path reports switch by its congestion information separately, and therefore, controller is after getting optimal path, for on optimal path be in congested switch the time obviously as can be known, if on optimal path, there is switch to be in congestion state, controller this optimal path as can be known is in congestedly, the data flow of input is divided into to two-way and forwards.

Step 207, according to above-mentioned forwarding strategy structure stream table list item, and send to corresponding switch, makes described switch get stream table list item, and storage.Wherein, the source node switch that described corresponding switch can be optimal path, also comprise switches all in network, when switch is not the source node switch of optimal path, this switch is not carried out the task of streamed data stream, flows to down hop but carry out forwarding data.In the present embodiment, the source node switch of optimal path, mainly, according to the stream table list item of controller structure, carry out the operation that streamed data flows and forwards, and other switches are only carried out and forwarded operation.

By above-mentioned explanation to the embodiment of the present invention two, this scheme middle controller is according to from each switch network, obtaining LSA message and NCA message, construct overall network topology figure and global congestion information, according to overall network topology figure and global congestion information, construct switch at two paths that are under congestion state, be about to congested data flow and be divided into two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Preferably, in above-mentioned steps 206, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, also specifically comprises:

Step 206a, the speed v of the data flow received in the source node switch of described optimal path is less than the maximum rate of influx that described optimal path allows, described data flow be take to speed and be the maximum rate of influx that described optimal path was allowed, forward according to the described optimal path obtained.

The forwarding strategy that the controller of above-mentioned increase generates, and make stream table list item according to forwarding strategy, its effect is when congestion condition does not appear in switch, just according to the optimal path transmitting data flow formed according to Global Topological figure.

Preferably, it should be noted that, data stream type can comprise: transmission control protocol (TCP, Transmission Control Protocol) data flow and User Datagram Protoco (UDP) (UDP, User Datagram Protocol) data flow.Wherein, UDP message stream is mainly real-time categorical data stream, and the characteristics of this class business are higher to requirement of real-time, relatively low for the requirement of packet loss, and can not carry out retransmission process for the packet abandoned; Adopt the data flow of Transmission Control Protocol transmission to be mainly the data class data flow, this class business is had relatively high expectations for packet loss, relatively low to requirement of real-time, packet loss is present if occur in network, for the packet abandoned, all can carry out retransmission process, if the frequent packet loss of such Business Stream, will cause a large amount of data packet retransmissions, cause the network congestion state further to worsen.Therefore, the forwarding strategy obtained in above-mentioned steps 206 specifically comprises:

Step 206b, described according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, the forwarding strategy when obtaining the data flow received in each switch and being tcp data stream.Wherein, described each switch comprises the source node switch of optimal path and the switch on each paths, but the source node switch that mainly refers to optimal path of indication streamed data stream in stream table list item in switch, switch in the middle of path is only carried out the forwarding task, does not carry out the operation of streamed data stream.Follow-up understanding to switch has similar explanation, does not repeat one by one.

Further, if when the data flow received in switch was UDP message stream, described method also comprised:

Receive the streaming rate that enters and leave each switch that each switch sends;

After step 203, this controller can also be carried out following steps:

Step 208, when the data flow received in corresponding switch is UDP message stream, according to the described streaming rate that enters and leave each switch, the drop probabilities ρ of construction data bag _iStrategy;

Step 209, according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding switch is UDP message stream;

Wherein, in step 208, when the data flow received in corresponding switch is UDP message stream, according to the described streaming rate that enters and leave each switch, the drop probabilities ρ of construction data bag _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding switch the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave switch i that switch reports, q _iFor the queue length of switch i, Q _ThFor the threshold value length of switch i, Q _maxMaximum permission queue length for switch i;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding switch is UDP message stream, specifically comprises:

When the queue length of switch was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of switch is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and switch that allows and be greater than while leaving speed, with The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of switch is greater than its maximum permission queue length, all UDP message bags that enter described switch all will be dropped.

Wherein, it will be appreciated that controller can generate corresponding stream table list item, stream table list item is handed down to respective switch by controller, stream table list item comprises three parts: head, counter and action, the header information of the direct matched data stream of switch, match direct execution and move accordingly, and this action is specially according to drop probabilities ρ _iPacket loss.

Step 210, the forwarding strategy when described data flow when receiving in corresponding switch is UDP message stream is made as stream table list item, sends to the switch of described correspondence.

It will be appreciated that said corresponding switch in step 208,209,210, refer to the switch on optimal path, after each switch on path receives UDP message stream, according to this switch congestion recorded in stream table list item, and according to the drop probabilities packet loss recorded; If this switch recorded in stream table list item is not congested, not packet loss; If record this switch heavy congestion, the packet that will receive all abandons.

By step 208, step 209, the step 210 of above-mentioned increase, make this controller to carry out different forwarding strategies to the UDP message stream of switch, namely with drop probabilities ρ _iStrategy forward.Making controller to carry out different forwardings to the different data streams in switch controls.

Embodiment tri-

The embodiment of the present invention provides a kind of method for routing of avoid congestion, the method receives mainly for the explanation switch stream table list item according to the forwarding strategy making that controller sends, thereby make when switch is follow-up receives data flow, the stream table list item that can inquire about storage carries out forwarding data flow.

The method comprises:

Step 301, receive the stream table list item that controller sends,

Step 302, store described stream table list item.

The method for routing of a kind of avoid congestion provided by the present embodiment, the method switch receive controller and send stream table list item, and when switch is follow-up while receiving data flow, the stream table list item that can inquire about storage carries out forwarding data flow.

Execute example four

The embodiment of the present invention provides a kind of method for routing of avoid congestion, the method is the angle explanation from switch, when having data flow to flow into switch, switch is specifically shunted according to the stream table list item data flow that controller sends, thereby effective avoid congestion, also avoid data re-transmission that packet loss brings etc. to cause network to continue the factor worsened.

As shown in Figure 7, the method comprises:

Step 401, the switch receiving data stream;

Step 402, the stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception,

Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item;

Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

The switch of this execution the method is that in optimal path, the source node switch is carried out triage operator, and the path intermediate node carries out according to the stream table list item that controller issues the operation that is forwarded to down hop, does not carry out the operation of shunting.

By above-mentioned explanation to the embodiment of the present invention four, in this scheme after the switch receiving data stream, according to the forwarding strategy recorded in stream table list item, congested data flow is divided into to two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Preferably, before the non-congestion state optimal path recorded in stream table list item when the data flow of being shunted forwards, expression (using ECN mechanism) can also be set at the packet head of this data flow of being shunted, make the destination node that receives this packet when the source node with sending to data flow is communicated by letter, inform that this source node sends streaming rate and causes too greatly congested, and then carry out the data flow adjustment from source node, thereby avoid congestion.Therefore, the method can also comprise:

Step 403, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, at the packet head of described another data flow, sign being set (can be specifically the 6th to be set to " 11 " for ECT position and the 7th CE position in IP head type of service field, wherein, the scheme that adopts ECT position and CE position to arrange mainly is based on ECN mechanism, ECN is the standard that IETF working group proposes, the congestion notification method of generally acknowledging at present, certainly, the present embodiment does not limit the situation of the setting of using one), for distinguishing the data flow that this data flow is congested shunting.

Preferably, in above-mentioned steps 402, can also comprise:

If the described optimal path recorded in described stream table list item is not to be in congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

Preferably, based on the controller side, UDP message stream and tcp data stream are issued respectively to not homogeneous turbulence table list item, therefore, for exchanger side, after step 402, the method also comprises:

Step 404, judge whether the data flow of described reception is tcp data stream,

If tcp data stream, execution step 403, the stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

Step 405, when the data flow of the described reception of judgement is not tcp data stream, carries out and forward according to the drop probabilities recorded in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described switch, and is less than the maximum queue length that allows; And the ingress rate of described switch is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave switch i, ρ _iDrop probabilities for switch i;

When the queue length of switch is greater than its maximum permission queue length, all UDP message bags that enter described switch all will be dropped.

Step 404 by above-mentioned increase～405, make switch can tackle different data flow and forward.

Preferably, after step 401, before step 402, the method can also comprise:

Step 406, judge whether described data flow is registered, if registration performs step 402;

Step 407, if unregistered, send to controller by the information of described data flow.

By increasing above-mentioned steps 406 to step 407, make switch more intelligent, to not reporting in time controller by packet out, processed by controller.

Embodiment five

The embodiment of the present invention provides a kind of method for routing of avoid congestion, and as shown in Figure 8 a, the method comprises:

Steps A 01, in network, each switch obtains information of neighbor nodes, according to information of neighbor nodes link generation state bulletin (LSA, the Link-State Advertisement) message obtained; And the LSA message is sent to controller.

It should be noted that, for dynamic network, the switch in network can be regularly to controller, send the LSA message, or current discovering neighbor nodal information sends the LSA message while changing, or other cause sending the LSA message in the situation of network change.

Wherein, in network, each switch obtains the concrete operations of information of neighbor nodes, can comprise: regularly peripherad neighbor node sends the Hello bag, obtain information of neighbor nodes, the link overhead that the switch measurement self is communicated by letter with each neighbor node is (as time delay, bandwidth etc.), become the LSA message to be uploaded to controller the information structuring of the neighbor node of collection, according to the LSA message of each switch, construct overall network topology figure by controller.As shown in Figure 9, each switch sends to the LSB message simplified schematic diagram of controller, and this figure is only the explanation of being convenient to understand, and is not the restriction to the embodiment of the present invention.As shown in Figure 9, in the LSA message that each switch sends, recorded each neighbours' of this switch information.

Step B02, controller receives the LSA (LSA, Link-State Advertisement) that in network, each switch sends; Controller, according to the LSA received, constructs overall network topology figure;

Steps A 03, in network, each switch obtains congestion state bulletin (NCA, Node Congestion Advertisement) message, and wherein, the NCA message is for showing whether this switch is in congestion state.The disposal ability of each switch is limited, and certain switch in network has been in congestion state, if now through the speed of this switch, surpass the speed of its deal with data, the congestion state of this node will further worsen so, in this programme, use P _iThe congested rate matrix that means node i, in network each switch regularly or when node state changes by p _iThe NCA message formed occurs to controller, and as shown in figure 10, the NCA message, comprise this node ID, and type of message generates reality, and the congested rate matrix P of node i _i.For controller, construct the global congestion rate matrix according to the NCA message of all nodes of collecting.

It should be noted that, the queue length threshold threshold value of supposing each switch is Q _Th, when the queue length of switch surpasses threshold value Q _ThThe time, switch can be understood as and is in congestion state.As shown in figure 11, in network, switch a is in congestion state, is just having two data to flow to into switch a simultaneously, and two data streams are respectively: λ _A1And λ _A2, the total quantity that enters so switch a is: λ _a(λ _a=λ _A1+ λ _A2), in like manner, also there are two data flow to process and open from switch a, the speed of leaving is respectively: μ _A1And μ _A2, total speed of leaving so switch a is μ _a(μ _a=μ _A1+ μ _A2).If now enter the speed of the data flow of switch a, be greater than the speed of leaving switch (that is: λ _aμ _a), this congestion state that just means node will further worsen, and now, we need to exceed node to this part and admit the data of ability to carry out extra processing, both needed to avoid node congested " overflowing " phenomenon to occur, also guaranteed that packet was not dropped and causes unnecessary re-transmission simultaneously.

P _iBe the congested rate matrix of definition, be used for meaning the congestion information of switch i, shown in (2):

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,i}\end{array}\right]}---\left(2\right)$

P _iBe the two-dimensional matrix of n * 1, n means the number of switch in zone, p _j,iMean that data flow allows ingress rate from the maximum that switch j enters switch i.It will be appreciated that, reality has exceeded the receptible scope of switch i from the ingress rate that switch j enters switch i, if continue the data access of current reality, will send so congested, so switch i notification controller is the speed after calculating, namely the controller speed that temporarily allows to enter switch i can not surpass p _j,i.

Wherein, in switch, can obtain p _j,iValue be as shown in the formula (3):

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.---\left(3\right)$

Wherein, λ _j,iExpression enters the streaming rate of switch i, λ from switch j _iMean all speed sums that enter switch i, value is

u _iMean all streaming rate sums of leaving switch i, value is

q _iMean the queue length of switch, work as q _iBe greater than the threshold value Q of switch length _Th, the speed that enters simultaneously switch i is greater than again the speed of leaving, and means that so the congestion state of switch i will further worsen, and all maximum rates that now allow to enter switch i from switch j are

For remaining

Data flow need to shunt processing.If now the switch flow velocity that is not in congestion state or enters switch is less than the speed of leaving, so, all data flow all will be walked according to normal routed path.

Step B04, controller receives the NCA message that in network, switch sends, and constructs the global congestion rate matrix, and following formula is shown in (4),

P=[P ₁ ... P _i ... P _n] （4）

Step B05, controller is according to overall network topology figure and global congestion rate matrix, obtain each node and go to the optimal path of other nodes in network and the maximum rate of influx that this path allows, and obtain non-congested optimal path, thereby construct the forwarding strategy of each node.

It should be noted that, the detailed step of step B05 can comprise:

Step B0500: the LSA message that controller is uploaded according to each switch, construct overall network topology figure, then use dijkstra's algorithm to calculate the optimal path of each switch to other switches

S is the access switch that data flow enters Local Area Network, and D is the telephone net node that leaves Local Area Network, k _iTelephone net node for process in the middle of every paths.

It will be appreciated that, about the algorithm that calculates optimal path, be not limited to above-mentioned dijkstra's algorithm in the present embodiment, can also be other algorithms that can calculate optimal path, should not be construed the restriction to the embodiment of the present invention herein.

Step B0502: controller inquiry global congestion rate matrix P obtains outbound path

On allow the maximum rate flow through to be $p=\mathrm{Min}(p_{S,{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">k</figure-callout>}}_1},...,p_{k_{i-1},k_i},...,p_{k_{n-1},k_n},p_{k_n,D});$

Step B0504: controller by all nodes of congestion state that are in network from topological diagram, removing, construct new network topological diagram, as shown in figure 12, controller utilizes dijkstra's algorithm again, according to new network topological diagram, calculates the optimal path from the source node to the destination node of non-congested network $L_{S,{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">k</figure-callout>}}_1^{\&prime;},...,k_i^{\&prime;},...,k_n^{\&prime;},D}^{\&prime;};$

Step B0506: controller is according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, and structure stream table, and send to corresponding switch.

Wherein, the source node switch that described corresponding switch can be optimal path, also comprise switches all in network, when switch is not the source node switch of optimal path, this switch is not carried out the task of streamed data stream, flows to down hop but carry out forwarding data.In the present embodiment, the switch of follow-up explanation, mainly refer to the source node switch of optimal path, if, according to the stream table list item of controller structure, carry out the operation that streamed data flows and forwards.

Whether controller is determined to shunt in the execution of source node switch, for switch fabric all on this path flows accordingly the table list item and sends.The stream table list item of each switch notifies this switch corresponding data flow is sent to which next-hop node.The list item of stream table as shown in figure 13 in the switch of source node S, for the data flow of going to destination node D, after stream table coupling, if the speed v of this data flow is less than the maximum in this path and allows speed p, this stream list item can inform that switch is by the optimal path of this data flow according to full meshed network topological diagram structure

By Packet Generation to next-hop node k ₁If the speed v of this data flow is greater than the maximum in this path and allows speed p, this stream list item can inform that switch is divided into two sub data flows by this data flow, the speed of a sub data flow allows the maximum rate p of access for this path, by the optimal path of this sub data flow according to full meshed network topological diagram structure Be sent to next-hop node k ₁, the sub data flow that a remaining speed is v-p will be according to the optimal path of non-congested network topological structure

Be sent to down hop k ' ₁.

Step B06, controller, according to the forwarding strategy of each node of structure, generate stream table list item, and send to corresponding switch.

By above-mentioned steps A01, steps A 03 step, make each switch in network that the LSA of self and NCA are sent to controller, by the foundation of controller as structure stream table.

Controller is by carrying out above-mentioned steps B02, B04, B05, B06, the LSA that can send according to each switch in network, the tectonic network topological diagram, NCA according to each switch transmission in network, structure global congestion rate matrix, thereby according to overall network topology figure and global congestion rate matrix, obtain each node and go to the optimal path of other nodes in network and the maximum rate of influx that this path allows, and obtain non-congested optimal path, finally construct the forwarding strategy of each node.This forwarding strategy makes in the stream table list item finally be stored in switch, include non-congested optimal path, make when data flow just enters network, with regard to the data-flow computation for exceeding the path access capability, go out the non-congested path of an optimum, rather than occurring under congestion situation, by switch, shunted or the packet loss processing, so both avoid network congestion, taken full advantage of again Internet resources, improved network throughput, simultaneously, also make load more balanced.

Preferably, above-mentioned steps B05, the forwarding strategy of each node gone out of controller structure, according to forwarding strategy structure stream table list item, can be not need to consider data stream type, wherein, data stream type can comprise: transmission control protocol (TCP, Transmission Control Protocol) data flow and User Datagram Protoco (UDP) (UDP, User Datagram Protocol) data flow.Namely do not distinguish data stream type, all carry out the operation as above-mentioned step B05.Further, if in controller, distinguish data stream type, the method also comprises:

Step B07: when data flow was tcp data stream, structure was if constructed the forwarding strategy of each node in step B05;

Step B08: when data flow is UDP message stream, controller is according to overall network topology figure and global congestion rate matrix, obtain each node and go to the optimal path of other nodes in network and the maximum rate of influx that this path allows, obtain drop probabilities ρ, wherein, contrast probability ρ abandons according to drop probabilities for the packet that will exceed the path receiving ability, thereby constructs the forwarding strategy of each node.

Wherein, the value of UDP message stream drop probabilities ρ is shown in formula (5):

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.---\left(5\right)$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave switch i, q _iFor the queue length of switch i, Q _ThFor the threshold value length of switch i, Q _maxMaximum permission queue length for switch i.When the queue length of switch was less than threshold value length, all packets all were not dropped; When the queue length of switch is more than or equal to threshold value length, is less than again the maximum ingress rate that allows queue length and switch and is greater than while leaving speed, with

The probability packet loss; When the queue length of switch is greater than its maximum permission queue length, so all UDP message bags that enter network all will be dropped, and according to probability ρ, carry out packet loss and can guarantee that the flow velocity that enters node is no more than its disposal ability.

As in step B08 when data flow be UDP message when stream, controller, before obtaining forwarding strategy, also needs to get the λ that switch reports _i, u _i, namely enter and leave the streaming rate of switch i, thereby can carry out the calculating as the described drop probabilities of step B08.

As shown in figure 14, the list item of the stream table that step B07 and step B08 obtain, in order to realize the differentiation to different business, during newly arrived Data stream query stream table, also will mate connection type (TCP, UDP) except mating purpose IP address.If there at certain switch ports themselves place, have a plurality of data flow to wait for to be processed, for the ageing while that guarantees UDP message stream is also avoided the packet loss of such business as far as possible, the UDP business can be set higher priority is arranged, switch, by priority treatment UDP message stream, makes it use as far as possible original shortest path.

It will be appreciated that, in network, UDP message stream is mainly real-time class data flow, and the characteristics of this class business are exactly higher to requirement of real-time, relatively low to the requirement of packet loss, and also can not carry out retransmission process for the packet abandoned; Adopt the data flow of Transmission Control Protocol transmission generally to be mainly the data class data flow, this class business is generally had relatively high expectations to packet loss, relatively low to requirement of real-time, if the packet loss phenomenon in network, occurs, for the packet abandoned, all can carry out retransmission process, if the frequent packet loss of such Business Stream, will cause a large amount of data packet retransmission phenomenons, so not only easily further worsen the network congestion state, also can be traffic impacting ageing, therefore, for these two kinds of different Business Streams, the technical program is taked the different disposal method.

The stream table how following step explanation switch utilizes controller to send, shown in Fig. 8 b, specifically comprises:

Steps A 09, switch receives the stream table list item that controller sends.

Steps A 10, the switch receiving data stream, according to the action in stream table list item, data flow is divided into to two sub-forwarding data flows, wherein, the speed of a sub data flow allows the maximum rate p of access for this path, by the optimal path of this sub data flow according to full meshed network topological diagram structure

Be sent to next-hop node k ₁, the sub data flow that a remaining speed is v-p will be according to the optimal path of non-congested network topological structure

Be sent to down hop k ' ₁.

Steps A 12, the described optimal path recorded in described stream table list item is in non-congestion state, the optimal path of constructing according to full meshed network topological diagram

Be sent to next-hop node k ₁.

By the explanation of above-mentioned steps A09～A12, switch is according to the stream table of storage, when data flow enters network, the data flow that will exceed the path access capability is transferred to a non-congested optimal path, simultaneously, guarantee that the data flow do not exceed the path access capability transmits on original optimal path, thereby data flow is divided into to two-way, therefore, can avoid network congestion, take full advantage of again Internet resources, improve network throughput, simultaneously, can arrive the effect of equally loaded.

Preferably, in steps A 11, the sub data flow that a remaining speed is v-p will be according to the optimal path of non-congested network topological structure

Be sent to down hop k ' ₁Operation before, can also comprise:

Steps A 1100, switch is in the sub data flow of v-p to a remaining speed, will belong in the IP head of packet of this data flow to arrange to mean it is the sign of streamed data bag;

Wherein, in the IP head, arrange and mean it is that the sign of streamed data bag can be specifically: the 6th be set to " 11 " for ECT position and the 7th CE position in IP head type of service field;

Steps A 11 specifically comprises: by the data flow arranged in steps A 1100, according to the optimal path of non-congested network topological structure

Be sent to down hop k ' ₁.

By switch, carry out above-mentioned steps A1100, ECN territory in the packet that switch will be shunted (wherein, in ECN mechanism, used two, IP head, one is the 6th and cries the ECT position, one is the 7th and is CE, these two be combined be called the ECN territory) setting, receiving terminal is received and is set to ' 11 ' in the ECN territory (" 11 " mean the 6th and are set to 1, the 7th also is set to 1) packet, congestion state has appearred in network as can be known, then by the ECN-Echo position, be filled to be ' 1 ' ack msg bag notice source reduction transmission rate, thereby more effectively solve the network congestion problem.

Preferably, before steps A 11, namely according to the action in the stream table, before data flow was divided into to two sub-forwarding data flows, the method can also comprise:

Steps A 13: whether the data stream type that the switch judgement receives is tcp data stream, if execution step A11; If not tcp data stream, carry out A14;

A14 wherein: switch, according to the action in the stream table, receives and enters the data flow in this switch according to drop probabilities ρ.

Wherein, the computing formula of drop probabilities ρ is suc as formula in (5), illustrating:

(1), when the queue length of switch is less than threshold value length, all packets all are not dropped;

(2), be less than again the maximum ingress rate that allows queue length and switch and be greater than while leaving speed when the queue length of switch is more than or equal to threshold value length, with

The probability packet loss;

(3), it is maximum while allowing queue length when the queue length of switch is greater than, so all UDP message bags that enter network all will be dropped, and according to probability ρ, carry out packet loss and can guarantee that the flow velocity that enters node is no more than its disposal ability.

Due to, in network, UDP message stream is mainly real-time class data flow, and the characteristics of this class business are exactly higher to requirement of real-time, relatively low to the requirement of packet loss, and also can not carry out retransmission process for the packet abandoned, adopt the data flow of Transmission Control Protocol transmission generally to be mainly the data class data flow, this class business is generally had relatively high expectations to packet loss, relatively low to requirement of real-time, if the packet loss phenomenon in network, occurs, for the packet abandoned, all can carry out retransmission process, if the frequent packet loss of such Business Stream, will cause a large amount of data packet retransmission phenomenons, so not only easily further worsen the network congestion state, also can be traffic impacting ageing, therefore, the embodiment of the present invention is by increasing steps A 13, A14, can process respectively for these two kinds of different Business Streams, improved the efficiency of transmission of UDP message stream and tcp data stream.

Embodiment six

The embodiment of the present invention provides a kind of network element, and this network element can be switch, but is not limited to switch, and as shown in figure 15, described network element comprises: the first acquiring unit C01, the first transmitting element C02, second acquisition unit C03, and the second transmitting element C04;

Described the first acquiring unit C01, be used to obtaining information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained;

Described the first transmitting element C02, for sending to controller by the LSA message;

Described second acquisition unit C03, be used to obtaining congestion state bulletin NCA message, wherein, include the congestion information of described network element in described NCA message;

Wherein, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

Described the second transmitting element C04, for sending to described controller by described NCA message.

A kind of network element by above-mentioned the present embodiment six explanations, network element in this scheme in network is by the information of neighbor nodes of self and self congestion information, by LSA and NCA, send to controller respectively, by controller to each network element unified management in network, construct overall network topology figure, with global congestion information, make controller according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and described optimal path, the maximum rate of influx that described optimal path allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that described optimal path allows, with described non-optimal path when congested, obtain forwarding strategy, and make stream table list item according to forwarding strategy, and the stream list item is sent to corresponding network element, make described network element obtain stream table list item, and storage; Wherein, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

The speed v of the data flow received in described corresponding network element is greater than the maximum rate of influx that described optimal path allows, the data flow of described reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as described optimal path, forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with described optimal path forwards according to described non-optimal path when congested.

Therefore, the technical scheme of the present embodiment explanation, compared with prior art, avoid simply abandoning the packet over formation max threshold length if abandoned in the tail algorithm, also overcome at network and do not occurred when congested can cause the problem of the waste of Internet resources with regard to starting in advance packet loss, a kind of method of more rational avoid congestion is provided, thereby has guaranteed the quality of transfer of data, for the user provides more high efficiency service.

Embodiment seven

The embodiment of the present invention provides a kind of controller, and as shown in figure 16, described controller comprises:

The first receiving element D01, the first structural unit D02, the second receiving element D03, the 3rd acquiring unit D04, the 4th acquiring unit D05, the 5th acquiring unit D06, the 6th acquiring unit D07, the 7th acquiring unit D08, the 3rd transmitting element D09;

Described the first receiving element D01, the LSA message sent for receiving each network element of network;

Described the first structural unit D02, for according to described LSA message, construct overall network topology figure;

Described the second receiving element D03, the NCA message sent for receiving each network element of network;

Described the 3rd acquiring unit D04, for according to described NCA message, obtain global congestion information;

Described the 4th acquiring unit D05, for according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

Described the 5th acquiring unit D06, for according to described global congestion information, and described optimal path, obtain the maximum rate of influx that described optimal path allows;

Described the 6th acquiring unit D07, for according to described global congestion information, by all nodes of congestion state that are in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

Described the 7th acquiring unit D08, for according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in optimal path, having network element to be in congestion state, require the source node network element of described optimal path that the data flow of described reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as described optimal path, forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with described optimal path forwards according to described non-optimal path when congested;

Described the 3rd transmitting element D09, for the described forwarding strategy obtained is sent to corresponding network element, make described network element obtain stream table list item according to described forwarding strategy, and storage.

By above-mentioned explanation to the embodiment of the present invention seven, this scheme middle controller is according to from each network element network, obtaining LSA message and NCA message, construct overall network topology figure and global congestion information, according to overall network topology figure and global congestion information, construct network element at two paths that are under congestion state, be about to congested data flow and be divided into two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Preferably, in described the 7th acquiring unit D08, for according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, also specifically comprises:

The speed v of the data flow received in the source node network element of optimal path is less than the maximum rate of influx that described optimal path allows, described data flow be take to speed and be the maximum rate of influx that described optimal path was allowed, forward according to the described optimal path obtained.

Preferably, each the NCA message received in the second receiving element D03 comprises the congestion information of a network element, and the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state;

In the 3rd acquiring unit, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of network element in zone, P _iCongested rate matrix for described network element.

Preferably, in described the 5th acquiring unit D06, for according to described global congestion information, and described optimal path, obtain the maximum rate of influx that described optimal path allows, specifically comprise:

From global congestion information and in described optimal path, obtain previous dive ki-1 in described optimal path and arrive down hop k _iMaximum allow ingress rate

Obtain in described optimal path minimum maximum and allow ingress rate, the maximum rate of influx p allowed as described optimal path, namely

Wherein, S is the source node of described optimal path, and D is the destination node of described optimal path.

Preferably, in described the 7th acquiring unit D08, for according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, is specially:

Described according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding network element is tcp data stream.

Preferably, described controller also comprises:

Described the second receiving element, the streaming rate that enters and leave each network element also sent be used to receiving each network element;

The second structural unit D10 and the 4th transmitting element D11;

Described the second structural unit D10, be used at described the 4th acquiring unit according to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding network element is UDP message stream, according to the described streaming rate that enters and leave each network element, structure drop probabilities ρ _iStrategy;

Described the 4th transmitting element D11, for by the described data flow received when corresponding network element, being that the forwarding strategy in UDP message when stream is made as stream table list item, send to the network element of described correspondence.

Preferably, while in described the second structural unit D10, being UDP message stream for the data flow received when corresponding network element, according to the described streaming rate that enters and leave each network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave network element i of reported by network elements, q _iFor the queue length of network element i, Q _ThFor the threshold value length of network element i, Q _maxMaximum permission queue length for network element i;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding network element is UDP message stream, specifically comprises:

When the queue length of network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and network element that allows and be greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

Embodiment eight

The embodiment of the present invention provides a kind of network element, and described network element comprises: the 3rd receiving element, and memory cell;

Described the 3rd receiving element, the stream table list item sent be used to receiving controller,

Described memory cell, be used to storing described stream table list item.

By the network element that the present embodiment provides, receive the stream table list item that controller sends, thereby make when network element is follow-up receives data flow, the stream table list item that can inquire about storage carries out forwarding data flow.

Embodiment nine

The embodiment of the present invention provides a kind of network element, and this network element can be switch, but is not limited to switch.As shown in figure 17, described network element comprises: the 4th receiving element E01, and the first retransmission unit E03;

Described the 4th receiving element E01, for receiving data stream;

Described the first retransmission unit E03, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

By above-mentioned to embodiment of the present invention explanation just, in this scheme after the network element receiving data stream, according to the forwarding strategy recorded in stream table list item, congested data flow is divided into to two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Preferably, described network element also comprises: indicate unit E04,

Described sign unit E04, for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

Preferably, described network element also comprises: the second retransmission unit E05,

Described the second retransmission unit E05, if the described optimal path recorded for described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

Preferably, described network element also comprises: the second judging unit E06,

Described the second judging unit E06, for the stream table list item of described the first retransmission unit according to storage, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described the first retransmission unit, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

Preferably, described network element also comprises: the 3rd retransmission unit E07,

Described the 3rd retransmission unit E07, while for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave network element i, ρ _iDrop probabilities for network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

Preferably, described network element also comprises: the 3rd judging unit E08, and the 5th transmitting element E09;

Described the 3rd judging unit, after described the 4th receiving element receiving data stream, before whether the described network element of described the first judgment unit judges is in congestion state, judge whether described data flow is registered;

If registration, notify described the first judging unit to carry out and describedly judge whether described network element is in the operation of congestion state;

Described the 5th transmitting element, if for unregistered, send to controller by the information of described data flow.

Embodiment ten

The embodiment of the present invention provides a kind of network element, and this network element can be switch, but is not limited to switch, as shown in figure 18, comprises the memory 40, processor 41, input unit 43 and the output device 44 that are connected respectively on bus, wherein:

In memory 40, be used for storing the data from input unit 43 inputs, and can also store the information such as necessary file of processor 41 deal with data;

Described processor, for the information of neighbor nodes according to obtaining from input unit, according to the information of neighbor nodes link generation state bulletin LSA message obtained; From input unit, obtain congestion state bulletin NCA message, wherein, in described NCA message, include the congestion information of described network element;

Described output device, for sending to controller by the LSA message; Described NCA message is sent to described controller.

Preferably, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

Preferably, described output device, also enter and leave the numerical value of the streaming rate of described network element for transmission, and the described numerical value that enters and leave the streaming rate of described network element obtains structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

A kind of network element by above-mentioned the present embodiment explanation, network element in this scheme in network is by the information of neighbor nodes of self and self congestion information, by LSA and NCA, send to controller respectively, by controller to each network element unified management in network, construct overall network topology figure, with global congestion information, make controller according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and described optimal path, the maximum rate of influx that described optimal path allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that described optimal path allows, with described non-optimal path when congested, obtain forwarding strategy, and make stream table list item according to forwarding strategy, and the stream list item is sent to corresponding network element, make described network element obtain stream table list item, and storage; Wherein, according to the described optimal path obtained, the maximum rate of influx that described optimal path allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

The speed v of the data flow received in described corresponding network element is greater than the maximum rate of influx that described optimal path allows, the data flow of described reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as described optimal path, forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with described optimal path forwards according to described non-optimal path when congested.

Therefore, the technical scheme of the present embodiment explanation, compared with prior art, avoid simply abandoning the packet over formation max threshold length if abandoned in the tail algorithm, also overcome at network and do not occurred when congested can cause the problem of the waste of Internet resources with regard to starting in advance packet loss, a kind of method of more rational avoid congestion is provided, thereby has guaranteed the quality of transfer of data, for the user provides more high efficiency service.

Embodiment 11

The embodiment of the present invention provides a kind of controller, as shown in figure 19, comprises the memory 50, processor 51, input unit 53 and the output device 54 that are connected respectively on bus, wherein:

In memory 50, be used for storing the data from input unit 53 inputs, and can also store the information such as necessary file of processor 41 deal with data;

Described input unit 53, the LSA message sent for receiving each first network element of network; Receive the NCA message that in network, each first network element sends;

Described processor 51, for according to described LSA message, construct overall network topology figure; According to described NCA message, obtain global congestion information; According to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, obtains stream table list item according to described forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested;

Described output device 54, send to described the first network element for the described stream table list item that will obtain.

Preferably, in described processor, for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, also specifically comprises:

The speed v of the data flow received when source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

Preferably, each the NCA message received in described input unit comprises the congestion information of first network element, and the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

$P_i=\stackrel{1}{\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{j,i}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ p_{n,\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="HDA00003585496400011.png"\; state="\{\{state\}\}">i</figure-callout>}}\end{array}\right]}$

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$P_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j\&CenterDot;i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

In described processor, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁ ... P _i ... P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

Preferably, in described processor, for according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely

Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

Preferably, in described processor, for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, is specially:

Described according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding the first network element is tcp data stream.

Preferably, described input unit, the streaming rate that enters and leave each first network element also sent be used to receiving each first network element;

Described processor also for:

According to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

Forwarding strategy when described data flow when receiving in corresponding the first network element is UDP message stream, be made as stream table list item, sends to the first network element of described correspondence.

Preferably, while in described processor, being UDP message stream for the data flow received when corresponding the first network element, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and the first network element that allows and be greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

By above-mentioned explanation to the embodiment of the present invention, this scheme middle controller is according to from each network element network, obtaining LSA message and NCA message, construct overall network topology figure and global congestion information, according to overall network topology figure and global congestion information, construct network element at two paths that are under congestion state, be about to congested data flow and be divided into two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Embodiment 12

The embodiment of the present invention provides a kind of network element, and this network element can be switch, but is not limited to switch.As shown in figure 20, comprise the memory 60, processor 61, input unit 63 and the output device 64 that are connected respectively on bus, wherein:

In memory 60, be used for storing the data from input unit 63 inputs, and can also store the information such as necessary file of processor 61 deal with data;

Described input unit 63, for receiving data stream;

Described processor 61, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

By above-mentioned to embodiment of the present invention explanation just, in this scheme after the network element receiving data stream, according to the forwarding strategy recorded in stream table list item, congested data flow is divided into to two data flow transmission, article one, data flow can be born speed with original path, according to this original path transmission, another of shunting data flow is transmitted with the optimal path under non-congestion state.Thereby the prior art packet discard of avoiding cause the not good problem of efficiency of transmission, greatly improved efficiency of transmission.

Preferably, described processor, also for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, arrange sign at the packet head of described another data flow, be used to distinguishing the data flow that this data flow is congested shunting.

Preferably, described processor, if the described optimal path also recorded for described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

Preferably, described processor, also, at the stream table list item of described the first retransmission unit according to storage, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described processor, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

Preferably, described processor, while also for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iFor in described stream table list item, recording the drop probabilities of network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of above-described embodiment is to come the hardware that instruction is relevant to complete by program, this program can be stored in a computer-readable recording medium, and storage medium can comprise: ROM, RAM, disk or CD etc.

Method for routing and the device of above a kind of avoid congestion that the embodiment of the present invention is provided, be described in detail, applied specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment is just be used to helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (45)

1. the method for routing of an avoid congestion, is characterized in that, comprising:

The first network element obtains information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained, and the LSA message is sent to controller;

Described the first network element obtains congestion state bulletin NCA message, and described NCA message is sent to described controller, wherein, in described NCA message, includes the congestion information of described the first network element.

2. method according to claim 1, is characterized in that, the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state.

3. according to claim 1 or 2 described methods, it is characterized in that, described method also comprises:

Described the first network element sends the numerical value of the streaming rate that enters and leave described the first network element, and the described numerical value that enters and leave the streaming rate of described the first network element obtains structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

4. the method for routing of an avoid congestion, is characterized in that, comprising:

Receive the LSA message that in network, each first network element sends; According to described LSA message, construct overall network topology figure;

Receive the NCA message that in network, each first network element sends; According to described NCA message, obtain global congestion information;

According to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

According to described global congestion information, and the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows obtained;

According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, according to described forwarding strategy, obtains stream table list item, and send to corresponding the first network element;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the described optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow of described reception, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested.

5. method according to claim 4, is characterized in that, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, also specifically comprises:

The speed v of the data flow received in source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

6. according to claim 4 or 5 described methods, it is characterized in that, each NCA message comprises the congestion information of first network element, and the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

According to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁…P _i…P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

7. method according to claim 6, is characterized in that, described according to described global congestion information, and the described optimal path obtained, and obtains the maximum rate of influx that the described optimal path obtained allows, and specifically comprises:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely $p=\mathrm{Min}(p_{S,k_1},...,p_{k_{i-1},k_i},...,p_{k_{n-1},k_n},p_{k_n,D}),$ Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

8. according to the described method of claim 4 to 7 any one, it is characterized in that, the optimal path that described basis is obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, is specially:

According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in the first network element is tcp data stream.

9. method according to claim 4, is characterized in that, described method also comprises:

Receive the streaming rate that enters and leave each first network element that each first network element sends;

Described according to described overall network topology figure, to obtain after each node goes to the optimal path of other node in network, described method also comprises:

When the data flow received in the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

According to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in the first network element is UDP message stream;

Forwarding strategy when described data flow when receiving in the first network element is UDP message stream, be made as stream table list item, sends to the first network element of described correspondence.

10. method according to claim 9, is characterized in that, when described data flow when receiving in the first network element is UDP message stream, and according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length and is less than, maximumly allow queue length, and the ingress rate of the first network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

11. the method for routing of an avoid congestion, is characterized in that, comprising:

Receiving data stream; Stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception,

Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item;

Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

12. according to the described method of claim 11, it is characterized in that, described another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward, before, described method also comprises:

Speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

13. according to the described method of claim 11, it is characterized in that, described method also comprises:

If the described optimal path recorded in described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

14. according to claim 11 to the described method of 13 any one, it is characterized in that, described stream table list item according to storage, before the data flow of described reception was divided into to two data flow forwardings, described method also comprised:

Whether the data flow that judges described reception is tcp data stream, if tcp data stream is carried out described stream table list item according to storage, the data flow of described reception is divided into to two data flow forwardings.

15. according to the described method of claim 14, it is characterized in that, described method also comprises:

When the data flow of the described reception of judgement is not tcp data stream, according to record in described stream table list item, carry out and forward, specifically comprise:

The queue length of packet is more than or equal to threshold value length in described the first network element, and is less than and maximum allows queue length, and the ingress rate of described the first network element is greater than while leaving speed, with ρ _iThe probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iDrop probabilities for record the first network element i in described stream table list item;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

16. a network element, is characterized in that, described network element comprises: the first acquiring unit, the first transmitting element, second acquisition unit, and the second transmitting element;

Described the first acquiring unit, be used to obtaining information of neighbor nodes, according to the information of neighbor nodes link generation state bulletin LSA message obtained;

Described the first transmitting element, for sending to controller by the LSA message;

Described second acquisition unit, be used to obtaining congestion state bulletin NCA message, wherein, include the congestion information of described network element in described NCA message;

Described the second transmitting element, for sending to described controller by described NCA message.

17. according to the described network element of claim 16, it is characterized in that, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

18. according to claim 16 or 17 described network elements, it is characterized in that, described the second transmitting element, also for transmission, enter and leave the numerical value of the streaming rate of described network element, the described numerical value that enters and leave the streaming rate of described network element, obtain structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

19. a controller, is characterized in that, described controller comprises:

The first receiving element, the first structural unit, the second receiving element, the 3rd acquiring unit, the 4th acquiring unit, the 5th acquiring unit, the 6th acquiring unit, the 7th acquiring unit, the 3rd transmitting element;

Described the first receiving element, the LSA message sent for receiving each first network element of network;

Described the first structural unit, for according to described LSA message, construct overall network topology figure;

Described the second receiving element, the NCA message sent for receiving each first network element of network;

Described the 3rd acquiring unit, for according to described NCA message, obtain global congestion information;

Described the 4th acquiring unit, for according to described overall network topology figure, obtain the optimal path that each node is gone to other node in network;

Described the 5th acquiring unit, for according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows;

Described the 6th acquiring unit, for according to described global congestion information, by all nodes of congestion state that are in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested;

Described the 7th acquiring unit, for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, obtains stream table list item according to described forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested;

Described the 3rd transmitting element, send to described the first network element for the described stream table list item that will obtain.

20. according to the described controller of claim 19, it is characterized in that, in described the 7th acquiring unit for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, also specifically comprise:

The speed v of the data flow received when source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

21. according to the described controller of claim 19, it is characterized in that, each the NCA message received in the second receiving element comprises the congestion information of first network element, the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

In the 3rd acquiring unit, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁…P _i…P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

22. according to the described controller of claim 21, it is characterized in that, in described the 5th acquiring unit, for according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely $p=\mathrm{Min}(p_{S,k_1},...,p_{k_{i-1},k_i},...,p_{k_{n-1},k_n},p_{k_n,D}),$ Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

23. according to claim 19 to the described controller of 22 any one, it is characterized in that, in described the 7th acquiring unit for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, be specially:

Described according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding the first network element is tcp data stream.

24. according to the described controller of claim 19, it is characterized in that described the second receiving element, the streaming rate that enters and leave each first network element also sent be used to receiving each first network element;

Described controller also comprises:

The second structural unit and the 4th transmitting element;

Described the second structural unit, be used at described the 4th acquiring unit according to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

Described the 4th transmitting element, for by the described data flow received when corresponding the first network element, being the UDP message forwarding strategy in when stream, be made as stream table list item, sends to the first network element of described correspondence.

25. according to the described controller of claim 24, it is characterized in that, while in described the second structural unit, being UDP message stream for the data flow received when corresponding the first network element, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and the first network element that allows and be greater than while leaving speed, with The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

26. a network element, is characterized in that, described network element comprises: the 4th receiving element, and the first retransmission unit;

Described the 4th receiving element, for receiving data stream;

Described the first retransmission unit, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

27. according to the described network element of claim 26, it is characterized in that, described network element also comprises: indicate unit,

Described sign unit, for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

28. according to the described network element of claim 26, it is characterized in that, described network element also comprises: the second retransmission unit,

Described the second retransmission unit, if the described optimal path recorded for described stream table list item is in non-congestion state, forward described data flow according to the optimal path recorded in described stream table list item.

29. according to the described network element of claim 26 to 28 any one, it is characterized in that, described network element also comprises: the second judging unit,

Described the second judging unit, at the stream table list item of described the first retransmission unit according to storage, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described the first retransmission unit, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

30. according to the described network element of claim 29, it is characterized in that, described network element also comprises: the 3rd retransmission unit,

Described the 3rd retransmission unit, while for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iFor in described stream table list item, recording the drop probabilities of network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

31. a network element is characterized in that described network element comprises the input unit be connected respectively on bus, output device, memory and processor;

Described processor, for the information of neighbor nodes according to obtaining from input unit, according to the information of neighbor nodes link generation state bulletin LSA message obtained; From input unit, obtain congestion state bulletin NCA message, wherein, in described NCA message, include the congestion information of described network element;

Described output device, for sending to controller by the LSA message; Described NCA message is sent to described controller.

32. according to the described network element of claim 31, it is characterized in that, the congestion information of described network element is specifically by the congested rate matrix P of described network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that network element j enters network element i, wherein, network element i is the network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of network element i from network element j; λ _iMean all speed sums that enter network element i, value is

u _iMean all streaming rate sums of leaving network element i, value is

q _iMean the queue length of network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of network element length _ThThe time, described network element is in congestion state.

33. according to claim 31 or 32 described network elements, it is characterized in that, described output device, also for transmission, enter and leave the numerical value of the streaming rate of described network element, the described numerical value that enters and leave the streaming rate of described network element, obtain structure drop probabilities ρ when data flow is UDP message stream for controller _iStrategy.

34. a controller is characterized in that described controller comprises the input unit be connected respectively on bus, output device, memory and processor;

Described input unit, the LSA message sent for receiving each first network element of network; Receive the NCA message that in network, each first network element sends;

Described processor, for according to described LSA message, construct overall network topology figure; According to described NCA message, obtain global congestion information; According to described overall network topology figure, obtain the optimal path that each node is gone to other node in network; According to described global congestion information, and the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows obtained; According to described global congestion information, by all nodes that are in congestion state in network from topological diagram, removing the network topological diagram that obtains non-congestion state; According to the network topological diagram of non-congestion state, obtain non-optimal path when congested; According to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, obtains stream table list item according to described forwarding strategy;

Wherein, according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain forwarding strategy, specifically comprises:

If when in the optimal path obtained, having the first network element to be in congestion state, require source node first network element of the described optimal path obtained that the data flow of reception is divided into to two data flow, wherein, data flow be take the maximum rate of influx that speed allows as the described optimal path obtained, and forwards according to the described optimal path obtained; Another data flow is with the speed v of the data flow that receives, and the speed of the difference of the maximum rate of influx allowed with the described optimal path obtained forwards according to described non-optimal path when congested;

Described output device, send to described the first network element for the described stream table list item that will obtain.

35. according to the described controller of claim 34, it is characterized in that, in described processor for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, also specifically comprise:

The speed v of the data flow received when source node first network element of the described optimal path obtained is less than the maximum rate of influx that the described optimal path obtained allows, described data flow be take to speed and be the described maximum rate of influx that optimal path was allowed obtained, forward according to the described optimal path obtained.

36. according to the described controller of claim 34, it is characterized in that, each the NCA message received in described input unit comprises the congestion information of first network element, the congestion information of described the first network element is specifically by the congested rate matrix P of described the first network element _iMean;

Wherein, P _iFor the two-dimensional matrix of n * 1, n means the number of the first network element in zone, p _j,iMean that data flow allows ingress rate from the maximum that the first network element j enters the first network element i, wherein, the first network element i is the first network element of described method executive agent;

Wherein, p _j,iValue be as shown in the formula:

$p_{j,i}=\left\{\begin{array}{cc}\frac{{\mathrm{\&lambda;}}_{j.i}}{{\mathrm{\&lambda;}}_i}\&CenterDot;u_i,& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,q_i\&GreaterEqual;Q_{\mathrm{th}};\\ \&infin;,& \mathrm{else};\end{array}\right.$

Wherein, λ _j,iExpression enters the streaming rate of the first network element i from the first network element j; λ _iMean all speed sums that enter the first network element i, value is

u _iMean all streaming rate sums of leaving the first network element i, value is

q _iMean the queue length of the first network element, work as λ _i>=μ _i, and q _iBe more than or equal to the threshold value Q of the first network element length _ThThe time, described the first network element is in congestion state;

In described processor, for according to described NCA message, obtain global congestion information, specifically comprise:

The congestion information of the first network element comprised according to described NCA message, construct global congestion rate matrix P, that is:

P=[P ₁…P _i…P _n]

Wherein, n means the number of the first network element in zone, P _iCongested rate matrix for described the first network element.

37. according to the described controller of claim 36, it is characterized in that, in described processor, for according to described global congestion information, and the described optimal path obtained, obtain the maximum rate of influx that the described optimal path obtained allows, specifically comprise:

From global congestion information and in the described optimal path obtained, obtain previous dive k in the described optimal path obtained _I-1To down hop k _iMaximum allow ingress rate

Obtain in the described optimal path obtained minimum maximum and allow ingress rate, the maximum rate of influx p allowed as the described optimal path obtained, namely $p=\mathrm{Min}(p_{S,k_1},...,p_{k_{i-1},k_i},...,p_{k_{n-1},k_n},p_{k_n,D}),$ Wherein, S is the source node of the described optimal path obtained, and D is the destination node of the described optimal path obtained.

38. according to the described controller of claim 34 to 37 any one, it is characterized in that, in described processor for according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, with described non-optimal path when congested, obtain forwarding strategy, be specially:

Described according to the described optimal path obtained, the maximum rate of influx that the described optimal path obtained allows, and described non-optimal path when congested, obtain the forwarding strategy when the data flow received in corresponding the first network element is tcp data stream.

39. according to the described controller of claim 34, it is characterized in that described input unit, the streaming rate that enters and leave each first network element also sent be used to receiving each first network element;

Described processor also for:

According to described overall network topology figure, obtain after each node goes to the optimal path of other node in network, when the data flow received in corresponding the first network element is UDP message stream, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy;

Forwarding strategy when described data flow when receiving in corresponding the first network element is UDP message stream, be made as stream table list item, sends to the first network element of described correspondence.

40. according to the described controller of claim 39, it is characterized in that, while in described processor, being UDP message stream for the data flow received when corresponding the first network element, according to the described streaming rate that enters and leave each first network element, structure drop probabilities ρ _iStrategy, specifically comprise:

Take following formula as the data flow when receiving in corresponding the first network element the drop probabilities ρ when UDP message flows _iStrategy:

${\mathrm{\&rho;}}_i=\left\{\begin{array}{cc}0,& q_i<Q_{\mathrm{th}};\\ \frac{{\mathrm{\&lambda;}}_i-u_i}{{\mathrm{\&lambda;}}_i},& {\mathrm{\&lambda;}}_i\&GreaterEqual;{\mathrm{\&mu;}}_i,Q_{\mathrm{th}}\&le;q_i<Q_{\max };\\ 1,& q_i\&GreaterEqual;Q_{\max };\end{array}\right.$

Wherein, λ _i, u _iBe respectively the streaming rate that enters and leave the first network element i of described the first reported by network elements, q _iBe the queue length of the first network element i, Q _ThBe the threshold value length of the first network element i, Q _maxThe maximum that is the first network element i allows queue length;

Described according to described drop probabilities ρ _iStrategy, and the described optimal path obtained, obtain the forwarding strategy when the data flow received in corresponding the first network element is UDP message stream, specifically comprises:

When the queue length of the first network element was less than threshold value length, all UDP message bags all forwarded with the described optimal path obtained;

When the queue length of the first network element is more than or equal to threshold value length also, be less than the maximum ingress rate of queue length and the first network element that allows and be greater than while leaving speed, with The probability packet loss, remaining UDP message bag forwards with the described optimal path that obtains;

When the queue length of the first network element is greater than its maximum permission queue length, all UDP message bags that enter described the first network element all will be dropped.

41. a network element is characterized in that described network element comprises the input unit be connected respectively on bus, output device, memory and processor;

Described input unit, for receiving data stream;

Described processor, the stream table list item for according to storage, be divided into two data flow forwardings by the data flow of described reception; Wherein, data flow be take the maximum rate of influx that speed records in described stream table list item optimal path allows, and forwards according to the optimal path recorded in described stream table list item; Another data flow is with the speed v of the data flow of described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, according in described stream table list item, record non-congested the time optimal path forward.

42. according to the described network element of claim 41, it is characterized in that, described processor, also for described the first retransmission unit according to described stream table list item, record non-congested the time optimal path forward before, speed v for the data flow with described reception, the speed of the difference of the maximum rate of influx allowed with the optimal path recorded in described stream table list item, another data flow forwarded, packet head in described another data flow arranges sign, be used to distinguishing the data flow that this data flow is congested shunting.

Described processor, if the described optimal path also recorded for described stream table list item is in non-congestion state, forwards described data flow according to the optimal path recorded in described stream table list item 43., according to the described network element of claim 42, it is characterized in that.

44. according to the described network element of claim 41 to 43 any one, it is characterized in that, described processor, also for described the first retransmission unit according to the storage stream table list item, before the data flow of described reception is divided into to two data flow forwardings, judge whether the data flow of described reception is tcp data stream;

Described processor, also, for if tcp data stream is carried out described stream table list item according to storage, be divided into two data flow forwardings by the data flow of described reception.

45. according to the described network element of claim 44, it is characterized in that, described processor, while also for the data flow when the described reception of judgement, not being tcp data stream, carrying out and forward according to record in described stream table list item, specifically comprises:

The queue length of packet is more than or equal to threshold value length in described network element, and is less than the maximum queue length that allows; And the ingress rate of described network element is greater than while leaving speed, with

The probability packet loss, remaining UDP message bag forwards with the optimal path that records in stream table list item; Wherein, ρ _iFor in described stream table list item, recording the drop probabilities of network element i;

When the queue length of network element is greater than its maximum permission queue length, all UDP message bags that enter described network element all will be dropped.

Images