CN108989148A - A kind of relaying multipath flow allocation method that propagation delay time minimizes - Google Patents

Abstract

The invention discloses a relay multi-path flow distribution method for minimizing transmission time delay. Traditional traffic distribution methods directly distribute end-to-end traffic on network sub-flows, which will cause transmission delay performance degradation. In the case of non-independent paths, the method of the invention splits the paths according to the convergence node, and performs relay multi-path flow distribution. This method first collects information and splits paths, then performs network path modeling and quality assessment, transmits traffic distribution and calculates the queuing delay of the path, solves the traffic distribution with the smallest delay, couples the traffic distribution results, and makes the data packets reach the destination node The average transmission delay is minimized. From the perspective of improving live broadcast delay performance, the method of the present invention considers the existence of convergence nodes in multi-path transmission, monitors link information in real time, and performs relay multi-path traffic distribution to obtain the minimum transmission delay, avoiding the traditional traffic distribution method. Transmission performance degradation caused by sub-flows competing for shared link resources.

Description

A Relay Multipath Traffic Distribution Method with Minimized Transmission Delay

technical field

The invention belongs to the technical field of network communication, and in particular relates to a relay multi-path flow distribution method for minimizing transmission time delay.

Background technique

With the development of network technology and the continuous emergence of various new media and self-media, the video streaming industry has made great progress, and the industry has higher and higher requirements for live broadcast delay and interactivity. Interactive live broadcast technology is already a standard configuration in the live broadcast industry. Under normal circumstances, the delay is less than 800 milliseconds to be able to do some relatively high-frequency interaction in the live broadcast, such as talk shows and live connections. If it is increased to 400 milliseconds, there will be enough margin to resist network fluctuations and realize interactive live broadcast. Therefore, how to better improve network delay performance has become a key factor for real-time video applications to ensure the quality of user interaction experience and rank among the industry's leading levels.

With the development of multiple access technologies, terminal devices usually have multiple different network interfaces and support different access technologies. Under the condition of sufficient bandwidth, using a single network access is not conducive to resource utilization, and manual switching of access methods will cause instantaneous interruption of services. To solve these problems, a multipath transmission technology (Concurrent Multipath Transmission, CMT) has received extensive attention.

In the application process of multi-path transmission technology, the transmission mode of segmentation/aggregation is usually adopted. When there is a sink node, the end-to-end path from the source node to the destination node is not independent. Different paths may share some link resources, and there is a competitive relationship. . Existing traffic allocation methods usually only model and analyze end-to-end traffic allocation. Through traffic allocation, the delay difference between different paths can be minimized, the delay jitter of data packets can be reduced, and the problem of out-of-sequence can be alleviated; dynamically adjusted according to the status of sub-flows The traffic allocated to sub-flows maximizes the throughput; the user experience quality modeling and evaluation during the transmission process optimizes the user experience quality through traffic allocation.

However, when traffic is allocated to non-independent paths, the allocated traffic will affect the transmission of other sub-flows on the shared link, and traffic allocation in the original way will cause transmission performance degradation. Therefore, it is of great significance how to better allocate traffic for multi-path transmission to obtain the minimum average transmission delay when the paths are not independent. The present invention considers the situation that there are converging nodes in multi-path transmission, divides paths according to the converging nodes, and performs relay multi-path flow distribution, so as to minimize the average transmission delay of data packets reaching destination nodes.

Contents of the invention

The purpose of the present invention is to provide a relay multi-path flow distribution method with minimum transmission time delay for the deficiencies of the prior art.

In the case of non-independent paths, the method of the present invention splits the paths according to the aggregation node, and performs relay multi-path traffic distribution, so as to minimize the average transmission delay of the data packet reaching the destination node, and the specific steps are as follows:

Step 1. Information Collection:

Monitor network traffic, collect and estimate link information; generate a set of paths from the source node to the destination node through the sink node according to the collected link information;

Think of the network as a directed graph G=(V,E), where Represents the collection of nodes, E={e _ij }, represents the collection of links between nodes; s represents the source node, d represents the destination node, n _ij represents the jth network node on the i-th path, N represents a general network node, N={1,2,...}, e _ij represents the j-th network link on the i-th path; the simple cycle-free path set between the source node s and the destination node d is P, P={P ₁ ,P ₂ ,...,P _K }, K is the number of paths.

Step 2. Path splitting:

Split the path from the source node s to the destination node d into multiple parts according to the sink node c, and redefine the logical path set of each part, P={P′:P″:...}={(P ₁ ′ ,P ₂ ′,P ₃ ′,P ₄ ′,...):(P ₁ ″,P ₂ ″,...):...}, where P′ represents the first logical path after splitting Set, P″ represents the second logical path set after splitting, P _i ′={e′ _i1 , e′ _i2 ,...}, P _i ″={e″ _i1 , e″ _i2 ,... } represents the mutually independent links between nodes on the i-th path, e′ _i1 represents the first link on the i-th path in the first logical path set, and e″ _i1 represents the second logical path set , the first link on the i-th path.

Step 3. Network path modeling:

Select a set of logical paths, obtain and update path network parameters: packet loss rate on path _Pi Indicates the packet loss rate on the link e _ij , the maximum available bandwidth a _i , the propagation delay pd _i , and the average transmission rate r _i , define the transmission available bandwidth on the path P _i w _i = _{ri +a i ;} define the path P trend bandwidth on _i For the current time t and the forecast value of the change trend of available bandwidth transmission at the previous q time: Among them, the parameters φ ₁ , φ ₂ , φ ₃ ..., φ _q are autoregressive coefficients, and ε _t is an independent white noise sequence.

Step 4. Quality assessment:

According to the available transmission bandwidth w _i on the path P _i and the packet loss rate p _i , combined with the time series model, calculate the evaluation bandwidth after quality evaluation: in and θ are weighting coefficients, respectively representing the weight of the packet loss rate and the weight of the trend bandwidth, 0<θ<1, satisfy

Step 5. Transmit traffic distribution:

The average rate at which all data packets arrive at the sender is λpackets/second. After reaching the source node, the sender assigns the data packets to K paths for transmission; each data packet is assigned to the i-th path with probability γ _i , and is Requested data packets arrive at path P _i at rate γ _i λ for transmission.

Step 6. Calculate the queuing delay of the path _Pi , and solve the traffic distribution with the minimum delay:

According to the queuing theory, the average transmission delay of the path P _i is given: the average time for the source node to send data packets on P _i Construct the traffic allocation problem with the minimum transmission delay, and solve the optimal traffic allocation vector γ=(γ ₁ ,γ ₂ ,...,γ _K ); λ _i represents the rate assigned to the i-th path, Indicates the evaluation bandwidth on the i-th path, and pd _i indicates the propagation delay on the i-th path;

Constraint C1 restricts the sending rate of the sender on each path to not exceed the maximum available bandwidth, and constraint C2 is a normative and non-negativity requirement for data packet allocation;

Define the Lagrange function μ, v, α are Lagrangian multipliers, Solve according to KKT conditions:

Where m is the number of paths selected for traffic allocation in the path set, and the traffic allocated on each path is:

Step 7. If the propagation delay difference of each path is less than the set time, it is considered that the propagation delay is similar, and proceed to step 8; if the propagation delay difference of each path is greater than or equal to the set time, it is considered that the propagation delay difference is relatively small. Large, go to step 9; the set time is 3-8 milliseconds;

Step 8. The propagation delays of the sub-flows of each path are similar, and the closed-form solution of the flow distribution is solved:

Go to step 10;

Step 9. The propagation delays of the sub-flows of each path are quite different, use the binary search to determine the upper and lower bounds of the search, and find the α approximate solution Get the flow distribution result; the details are as follows:

Step 9.1. Set the search accuracy σ, and determine the upper and lower bounds of the binary search:

Step 9.2. Update the median value of the binary search

Step 9.3. Calculate the verdict:

like Adjust the search lower bound, Return to step 9.2; if Adjust the upper bound of the binary search, Return to step 9.2; if Get an approximate solution with accuracy σ

Step 9.4. The obtained substitute Get the traffic distribution result:

Step 10. If there is a logical path set without traffic allocation, perform traffic allocation on the next logical path set, and go to step 3; otherwise, go to step 11.

Step 11. Coupling Traffic Distribution Results:

Coupling the traffic distribution results of each part, generating the transmission path P={P′:P″:...} and traffic distribution {γ′+γ″,...} results from the source node to the destination node, and sending data ; If a new data packet arrives after the transmission is completed, enter step 3, and perform relay multi-path traffic distribution with minimum transmission delay for a new round of data transmission; otherwise, end and exit.

From the perspective of improving live broadcast delay performance, the method of the present invention considers the existence of convergence nodes in multi-path transmission, monitors link information in real time, and performs relay multi-path traffic distribution to obtain minimum transmission delay. Compared with traditional traffic distribution methods, its advantages are reflected in:

Traditional traffic distribution methods usually perform end-to-end traffic distribution directly on network sub-flows. When network subflows converge at intermediate nodes, the allocated traffic will affect the transmission of other subflows on the shared link, and traffic distribution in the original way will cause transmission performance degradation. However, the method of the present invention avoids the decline in transmission performance caused by sub-flow competition for shared link resources in the traditional flow distribution method through path splitting, transmission delay minimization, flow distribution, and coupling of flow distribution results, so that data packets can reach the destination. The average transmission delay of nodes is minimized.

Description of drawings

Fig. 1 is the flowchart of the inventive method;

Figure 2 shows a multi-path transmission network topology with a sink node.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings and examples, and the method flow is shown in FIG. 1 .

The present invention takes a converging node as an example to illustrate the relay multipath traffic distribution method, and the topology of the multipath transmission network is shown in FIG. 2 . There is a sink node c between the source node s and the destination node d. The transmission delay of each link is set to 10ms, the available bandwidth of the bottleneck link e′ ₁₁ , e′ ₂₁ , and e′ ₃₁ is set to 6Mbps, 4Mbps and 5Mbps respectively, and the packet loss rate is set to 0.1%. The available bandwidth is set to 10Mbps, and the packet loss rate is set to 0.

1. Information collection. Monitor network traffic, collect and estimate link information; generate a set of paths from the source node through the sink node to the destination node based on the collected link information. Think of the network as a directed graph G=(V,E), where Represents the collection of nodes, E={e _ij }, represents the collection of links between nodes; s represents the source node, d represents the destination node, n _ij represents the jth network node on the i-th path, N represents a general network node, N={1,2,...}, e _ij represents the j-th network link on the i-th path; the simple cycle-free path set between the source node s and the destination node d is P, P={P ₁ ,P ₂ ,P ₃ ,P ₄ ,P ₅ ,P ₆ }, the number of paths is 6.

2. Path splitting. Split the path from the source node s to the destination node d into two parts according to the sink node c, and redefine the logical path set of each part, P={P′:P″}={(P ₁ ′,P ₂ ′ ,P ₃ ′):(P ₁ ″,P ₂ ″)}, where P′ and P″ represent the logical path of each part after splitting, P _i ′={e′ _i1 ,e′ _i2 }, P _i ″ ＝{e″ _i1 ,e″ _i2 } represents the independent links among the nodes on the i-th path;

3. Network path modeling. Select the logical path set P′, obtain and update the network parameters on the path P _i :

Packet loss rate: p _i =0.1%, i=1,2,3;

Maximum available bandwidth: a ₁ =6Mbps, a ₂ =4Mbps, a ₃ =5Mbps;

Propagation delay: pd _i = 20ms, i = 1, 2, 3;

Average transmission rate: r _i = 0, i = 1, 2, 3;

Available bandwidth for transmission: w ₁ =6Mbps, w ₂ =4Mbps, w ₃ =5Mbps;

Trend Bandwidth:

4. Quality assessment. Take the packet loss rate weight Take the trend bandwidth weight θ=0.9 as an example, according to the transmission available bandwidth w _i and packet loss rate p _i on the path P _i , calculate the evaluation bandwidth after quality evaluation:

5. Transmission traffic distribution. Taking the sending rate of 12.32Mbps as an example, the size of each data packet is set to 1400 bytes, and the average rate of arrival of data packets is 1100 packets/second.

6. Calculate the queuing delay of the path P _i , and find the closed-form solution γ′=(γ′ ₁ , γ′ ₂ , γ′ ₃ ) of the logical path set P′ flow distribution.

7. The difference in the propagation delay pd _i of each path is small, go to step 8;

8. Obtain the closed-form solution γ' of logical path set P' flow distribution:

γ' ₁ =0.41, γ' ₂ =0.26, and γ' ₃ =0.33.

9. Repeat the above process for the logical path set P″, and obtain the flow distribution result γ″ of the logical path set P″, where γ″ ₁ =0.5, γ″ ₂ =0.5.

10. Coupling flow distribution results. Coupling the traffic distribution results of each part, generating the transmission path set P={P′:P″:...} and traffic distribution {γ′+γ″,...} results from the source node to the destination node, and performing data send:

P ₁ = {e′ ₁₁ , e′ ₁₂ , e″ ₁₁ , e″ ₁₂ }, γ ₁ = 0.205;

P ₂ = {e′ ₁₁ , e′ ₁₂ , e″ ₂₁ , e″ ₂₂ }, γ ₂ = 0.205;

P ₃ = {e′ ₂₁ , e′ ₂₂ , e″ ₁₁ , e″ ₁₂ }, γ ₃ = 0.13;

P ₄ ={e′ ₂₁ , e′ ₂₂ , e″ ₂₁ , e″ ₂₂ }, γ ₄ =0.13;

P ₅ = {e′ ₃₁ , e′ ₃₂ , e″ ₁₁ , e″ ₁₂ }, γ ₅ = 0.165;

P ₆ = {e′ ₃₁ , e′ ₃₂ , e″ ₂₁ , e″ ₂₂ }, γ ₆ = 0.165;

11. Go back to step 3, and perform relay multi-path traffic distribution with minimum transmission delay for a new round of data transmission until the transmission is completed.

12. If the propagation delay pd _i of each path is quite different, pd ₁ =20ms, pd ₂ =40ms, pd ₃ =30ms, use binary search to calculate traffic distribution.

13. Determine the upper and lower bounds Set search precision σ=1.

14. Update the intermediate value of the binary search

14. Calculation of judgments, Adjust the search lower bound,

16. Update the intermediate value of the binary search

17. Calculation of judgments, Adjust the search lower bound,

18. Repeat the binary search process until a satisfying Approximate solution of

19. According to the approximate solution For traffic distribution:

γ′ ₁ =0.4126, γ′ ₂ =0.2545, and γ′ ₃ =0.3329.

Claims (1)

1. the relaying multipath flow allocation method that a kind of propagation delay time minimizes, it is characterised in that this method specific steps are such as Under:

Step 1. information is collected:

Monitoring network flow is collected and estimates link information；According to the link information being collected into, generate from source node by converging Set of paths of the node to destination node；

Network is regarded as digraph G=(V, E), whereinIndicate the set of node, E={ e_ij, indicate section The set of link between point；S indicates that source node, d indicate destination node, n_ijIndicate that j-th of network node on the i-th paths, N indicate General networking node, N=1,2 ... }, e_ijIndicate j-th strip network link on the i-th paths；Source node s and destination node d it Between it is simple without circulation available path collection be combined into P, P={ P₁,P₂,...,P_K, K is number of path；

Step 2. path is split:

Source node s to the path of destination node d is split into multiple portions according to aggregation node c, and redefines various pieces Logical path set, P={ P ': P ": ... }={ (P₁′,P₂′,P₃′,P₄′,...):(P₁″,P₂" ...): ... }, wherein P ' Indicate first logical path set after splitting, P " indicates second logical path set after splitting, P_i'={ e '_i1,e ′_i2... }, P_i"={ e "_i1,e″_i2... } and indicate mutually indepedent link on the i-th paths between each node, e '_i1Indicate first In a logical path set, first link on the i-th paths, e '_i1It indicates in second logical path set, i-th road First link on diameter；

The modeling of step 3. network path:

A logical path set is chosen, obtain and updates path network parameter: path P_iOn packet loss Indicate link e_ijOn packet loss, maximum available bandwidth a_i, propagation delay pd_i, average transmission speed Rate r_i, define path P_iOn transmission available bandwidth w_i=r_i+a_i；Define path P_iOn trend bandwidthFor current time t with The predicted value of preceding q moment transmission available bandwidth variation tendency:Wherein parameter φ₁,φ₂,φ₃...,φ_qFor autoregressive coefficient, ε_tFor mutually independent white noise sequence；

Step 4. quality evaluation:

According to path P_iOn transmission available bandwidth w_i, packet loss p_i, binding time series model, commenting after calculating quality evaluation Estimate bandwidth:WhereinIt is weighting coefficient with θ, respectively indicates packet loss weight and trend bandwidth is weighed Weight,0 < θ < 1 meets

Step 5. transmits assignment of traffic:

The Mean Speed for the transmitting terminal that all data groupings reach is λ packets/second, and after reaching source node, transmitting terminal divides data Group is assigned on K paths and transmits；Each data grouping is with probability γ_iIt is assigned on the i-th paths, requested data grouping With rate γ_iλ reaches path P_iIt is sent；

Step 6. calculates path P_iQueuing delay, solve the smallest assignment of traffic of time delay:

According to queueing theory, path P is provided_iMean transit delay: source node is in P_iThe upper average time for sending data groupingConstruct the smallest assignment of traffic problem of propagation delay time, and solve optimize assignment of traffic vector γ= (γ₁,γ₂,...,γ_K)；λ_iIndicate the rate for being assigned to the i-th paths,Indicate the assessment bandwidth on the i-th paths, pd_iTable Show the propagation delay on the i-th paths；

Constraint condition C1 limits transmission rate of the transmitting terminal in each path no more than maximum available bandwidth, constraint condition C2 It is the normalization and nonnegativity requirement to data grouping distribution；

Define Lagrangianμ, v, α are Lagrange Multiplier,It is solved according to KKT condition:

Wherein m is the path number that assignment of traffic is selected in set of paths, the flow distributed on each paths are as follows:

If the propagation delay difference of each paths of step 7. is less than setting time, it is close to be considered as propagation delay, enters step 8；If each The propagation delay difference of paths is more than or equal to setting time, is considered as propagation delay and differs larger, enters step 9；The setting Time is 3~8 milliseconds；

The propagation delay of each paths subflow of step 8. is close, solves the closed solutions of assignment of traffic:

Enter step 10；

Step 9. uses binary search, determines the bound of search, seeks α approximate solutionAssignment of traffic is obtained as a result, specific as follows:

Search precision σ is arranged in step 9.1., determines the bound of binary search:

The median of step 9.2. update binary search

Step 9.3. calculates judgement:

IfAdjustment search lower bound,It returns Step 9.2；IfThe upper bound of binary search is adjusted,Return step 9.2；IfAcquire precision For the approximate solution under σ

Step 9.4. will be acquiredIt substitutes intoObtain assignment of traffic result:

Step 10. does not carry out the logical path set of assignment of traffic if it exists, then carries out flow to next logical path set Distribution, enters step 3；Otherwise, 11 are entered step；

Step 11. coupling traffic allocation result:

Each section assignment of traffic result is coupled, generate source node to destination node transmission path P={ P ': P ": ... } With assignment of traffic { γ '+γ " ... } as a result, carrying out data transmission；If there is new data grouping to reach after being sent completely, enter Step 3, new round data are transmitted and carries out the smallest relaying multipath assignment of traffic of propagation delay time；Otherwise, terminate and exit.