CN1604570A - Flow distribution method based on numerical differentiation - Google Patents

Abstract

This invention relates to a flow alignment method based on numerical value differentiation and belongs to information transmission technique field. In this method, n pieces of flow media data source sends the flow media data in form of RTP message, each RTP message has one sequence and realizes the flow media data alignment through division of RTP message sequence. The method is based on value differentiation and refers to line scan conversion in reference graphics and converts flow alignment into a sequence split math model and realizes accurate flow alignment.

Description

A kind of flow allocation method based on numerical differentiation

Technical field

The invention belongs to the information dissemination technology field, relate in particular to a kind of assignment of traffic design in the stream scheduling method under the multi-data source network architecture environment that is applied to based on numerical differentiation.

Background technology

In the network architecture of a multi-data source, a plurality of data source of stream media provide stream medium data for a client node simultaneously.In order to reach the purpose of stream scheduling, must there be an effective flow allocation method to support, different flow allocation methods will directly have influence on final stream dispatching effect.

In the document transmission process of present multithreading,, can't realize intelligence such as network characteristic, load capacity and transmission performance, dispense flow rate accurately according to each data source though can realize the file transfer of multi-data source.Simultaneously, traditional file transfer all is file to be carried out segmentation split, in case data packet loss phenomenon takes place, may cause the continuous packet loss that transmits, carries out data and recovers thereby have a strong impact on error correcting system such as forward error correction.So far, traditional flow allocation method distributes at intelligent flow, packet loss is handled and all there is defective in the data dependence direction.

Summary of the invention

The objective of the invention is to propose a kind of flow allocation method based on numerical differentiation.Linear scanning conversion during this method reference pattern is learned can realize comparatively accurate assignment of traffic.

A kind of flow allocation method that the present invention proposes based on numerical differentiation, it is characterized in that, n data source of stream media sent stream medium data with the form of RTP message, each RTP message all has a sequence number (sequencenumber), realizes distribution to the streaming media data stream amount by the division to the RTP test serial number; Specifically may further comprise the steps:

(1) the RTP sequence of message is defined as a sequence { a _m=m+S ₀, m 〉=0 ∧ m ∈ Z, in the formula, S ₀Start sequence number for this sequence;

(2) the inhomogeneous allocation table that brings of the heterogeneous networks load capacity of this n data source of stream media is shown the ratio B of each subsequence element number that this sequence obtains after dividing ₁: B ₂: ...: B _n, the subsequence of the n after the division is $\left\{a_k^{\left(1\right)}\right\},\·\·\·,\left\{a_k^{\left(n\right)}\right\},k\∈Z^{+}\∪\left\{0\right\},$ Wherein n is the number (the corresponding subsequence of data source of stream media) of subsequence;

(a 3) i subsequence { a _k ⁽ⁱ⁾It is the sequence number of corresponding i the data source of stream media RTP message that need send to client node.

In the said method, the detailed method that sequence is divided comprises the steps:

(2.1) determine initial test serial number S ₀, subsequence number n, the ratio B of subsequence element number ₁: B ₂: ...: B _n, and the sequence number i that the client node of Media Stream is provided;

(2.2) inequality 2 is satisfied in calculating ^L-1＜max (B ₁..., B _n)＜2 ^lL;

(2.3) variable assignments dt=2 ^-l, ds ₁=B ₁Dt, ds ₂=B ₂Dt ..., ds _n=B _nDt;

(2.4) variable assignments s ₁=s ₂=...=s _n=0, m=S ₀, k=0;

(2.5) when satisfying k＜k _Max(k _MaxThe length of representing i subsequence) time, execution in step (2.6) is to (2.11);

(2.6) for each j from 1 to n, execution in step (2.7) is to (2.11);

(2.7) if s _j+ ds _j≠ [s _j], execution in step (2.8) is to (2.10);

(2.8) if j=i, execution in step (2.9);

(2.9) variable assignments $a_k^{\left(i\right)}=m,$ k＝k+1；

(2.10) variable assignments m=m+1;

(2.11) variable assignments s _j=s _j+ ds _j

(2.12) obtain i subsequence $\left\{a_k^{\left(i\right)}\right\},0<k<k_{\max },k\&Element;Z.$

Characteristics of the present invention and effect:

The inventive method is based on numerical differentiation, and the linear scanning conversion during reference pattern is learned converts the assignment of traffic problem to Mathematical Modeling that a sequence splits, realizes comparatively accurate assignment of traffic.This method breaks through existing segmentation method for splitting, splits by the great-jump-forward sequence, has reduced contingent continuous drop probabilities effectively, recovers to have created condition for error correcting systems such as forward error correction carry out data.

Description of drawings

Fig. 1 is the flow allocation method flow chart in the inventive method

Fig. 2 is the flow allocation method embodiment schematic diagram of the inventive method when two data source of stream media.

Embodiment

A kind of flow allocation method based on numerical differentiation that the present invention proposes reaches embodiment in conjunction with the accompanying drawings and is described in detail as follows:

Flow allocation method based on numerical differentiation of the present invention as shown in Figure 1, it is characterized in that, n data source of stream media sent stream medium data with the form of RTP message, each RTP message all has a sequence number (sequencenumber), realizes distribution to the streaming media data stream amount by the division to the RTP test serial number; Specifically may further comprise the steps:

(1) the RTP sequence of message is defined as a sequence { a _m=m+S ₀, m 〉=0 ∧ m ∈ Z, in the formula, S ₀Start sequence number for this sequence;

(2) the inhomogeneous allocation table that brings of the heterogeneous networks load capacity of this n data source of stream media is shown the ratio B of each subsequence element number that this sequence obtains after dividing ₁: B ₂: ...: B _n, the subsequence of the n after the division is $\left\{a_k^{\left(1\right)}\right\},\&CenterDot;\&CenterDot;\&CenterDot;,\left\{a_k^{\left(n\right)}\right\},k\&Element;Z^{+}\&cup;\left\{0\right\},$ Wherein n is the number (the corresponding subsequence of data source of stream media) of subsequence;

(a 3) i subsequence { a _k ⁽ⁱ⁾It is the sequence number of corresponding i the data source of stream media RTP message that need send to client node.

In the said method, the detailed method that sequence is divided comprises the steps:

(2.1) determine initial test serial number S ₀, subsequence number n, the ratio B of subsequence element number ₁: B ₂: ...: B _n, and the sequence number i that the client node of Media Stream is provided;

(2.2) inequality 2 is satisfied in calculating ^L-1＜max (B ₁..., B _n)＜2 ^lL;

(2.3) variable assignments dt=2 ^-l, ds ₁=B ₁Dt, ds ₂=B ₂Dt ..., ds _n=B _nDt;

(2.4) variable assignments s ₁=s ₂=...=s _n=0, m=S ₀, k=0;

(2.5) when satisfying k＜k _Max(k _MaxThe length of representing i subsequence) time, execution in step (2.6) is to (2.11);

(2.6) for each j from 1 to n, execution in step (2.7) is to (2.11);

(2.7) if s _j+ ds _j≠ [s _j], execution in step (2.8) is to (2.10);

(2.8) if j=i, execution in step (2.9);

(2.9) variable assignments $a_k^{\left(i\right)}=m,$ k＝k+l；

(2.10) variable assignments m=m+1;

(2.11) variable assignments s _j=s _j+ ds _j

(2.12) obtain i subsequence $\left\{a_k^{\left(i\right)}\right\},0<k<k_{\max },k\&Element;Z.$

The flow allocation method embodiment of the inventive method when two data source of stream media as shown in Figure 2, flow allocation method can be described as linear scanning conversion (calculating process of exploded view 1 directly perceived) intuitively.In the present embodiment, subsequence element number n=2, the element number ratio B of two subsequences ₁: B ₂=10: 7, sequence initial sequence number S ₀=1, obtain two subsequences $\left\{a_k^{\left(1\right)}\right\},\left\{a_k^{\left(2\right)}\right\},k\&Element;Z^{+}\&cup;\left\{0\right\}.$

Demarcated two-dimensional coordinate system among Fig. 2, each grid (pixel) is all represented a pixel.Here put to straight dashed line of (10,7) strokes and dots by (0,0), and utilize some pixels (i.e. grey grid among the figure) of picture element matrix to approach this straight dashed line.

Among Fig. 2 by (0,0) goes to (10,7) point is (shown in the figure arrow, here requirement can only be advanced and preferential along the Y coordinate along the direction that the XY coordinate increases, also but the X coordinate is preferential), to advance altogether for 17 steps (certainly moving on) along this straight line, each step has all been indicated sequence number, wherein walked for 10 steps, walked for 7 steps along the Y direction along directions X.At last will be along the sequence number of directions X as subsequence 1, along the sequence number of Y direction as subsequence 2.Obtain two subsequences

According to this flow allocation method, the RTP test serial number that last data source of stream media 1 need send to client node is 2,3,5,7,9,10,12,14,15,17 ...; The RTP test serial number that data source of stream media 2 need send to client node is 1,4,6,8,11,13,16 ....

Because packet loss mostly is the continuous packet loss of burst greatly in the Network Transmission, if do not adopt the fractionation of this great-jump-forward RTP test serial number, but adopt simple segmentation to split, for example for above-mentioned sequence, its two sequences that are divided into 10: 7 are had: sequence 1 is 1～10,18～27,35～44, ... sequence 2 is 11～17,28～34,45～51, ..., will make the probability of continuous packet loss increase greatly, recover, cause expendable data segment to increase thereby be unfavorable for using error correcting system such as forward error correction to carry out data.And adopt this flow allocation method, and all messages are evenly broken up, can be so that the probability of the continuous packet loss that happens suddenly reduce greatly, thus make that the data recovered section does not significantly reduce.

Claims (2)

1, a kind of flow allocation method based on numerical differentiation, it is characterized in that, n data source of stream media sent stream medium data with the form of RTP message, each RTP message all has a sequence number, realizes distribution to the streaming media data stream amount by the division to the RTP test serial number; Specifically may further comprise the steps:

1) the RTP sequence of message is defined as a sequence { a _m=m+S ₀, m 〉=0 ∧ m ∈ Z, in the formula, S ₀Start sequence number for this sequence;

2) the inhomogeneous allocation table that brings of the heterogeneous networks load capacity of this n data source of stream media is shown the ratio B of each subsequence element number that this sequence obtains after dividing ₁: B ₂: ...: B _n, the subsequence of the n after the division is $\left\{a_k^{\left(l\right)}\right\},\&CenterDot;\&CenterDot;\&CenterDot;,\left\{a_k^{\left(n\right)}\right\},k\&Element;Z^{+}\&cup;\left\{0\right\}$ , wherein n is the number of subsequence, the corresponding subsequence of data source of stream media;

3) i subsequence { a _k ⁽ⁱ⁾It is the sequence number of corresponding i the data source of stream media RTP message that need send to client node.

2, in the said method as claimed in claim 1, it is characterized in that the described the 2nd) method that sequence in the step is divided, specifically comprise the steps:

(1) determines initial test serial number S ₀, subsequence number n, the ratio B of subsequence element number ₁: B ₂: ...: B _n, and the sequence number i that the client node of Media Stream is provided;

(2) inequality 2 is satisfied in calculating ^L-1＜max (B ₁..., B _n)＜2 ^lL;

(3) variable assignments dt=2 ^-l, ds ₁=B ₁Dt, ds ₂=B ₂Dt ..., ds _n=B _nDt;

(4) variable assignments s ₁=s ₂=...=s _n=0, m=S ₀, k=0;

(5) when satisfying k＜k _Max(k _MaxThe length of representing i subsequence) time, execution in step (2.6) is to (2.11);

(6) for each j from 1 to n, execution in step (2.7) is to (2.11);

(7) if s _j+ ds _j≠ [s _j], execution in step (2.8) is to (2.10);

(8) if j=i, execution in step (2.9);

(9) variable assignments $a_k^{\left(i\right)}=m,k=k+1;$

(10) variable assignments m=m+1;

(11) variable assignments s _j=s _j+ ds _j

(12) obtain i subsequence $\left\{a_k^{\left(i\right)}\right\},0<k<k_{\max },k\&Element;Z.$

Images