CN103986633A - Subnetting method based on 1394b multi-subnet transmission structure - Google Patents

Abstract

The invention provides a subnetting method based on a 1394b multi-subnet transmission structure and belongs to the technical field of network information communication. The method comprises the steps that a matrix A of transmission bandwidth between network nodes is obtained; a network topology matrix is given as X, each column of X is a subnet, and when a r<th> node is in a certain subnet, the value of a r<th> row element corresponding to the vector of the subnet is 1; all the network topology matrixes are enumerated; delay matrixes T in the subnets and between the subnets under the network topology are computed, and the average delay of a whole network is obtained; and according to the network topology matrix corresponding to the minimum value of the whole network average delay, subnetting is carried out. According to the method, 1394b multi-subnet structure dividing is achieved, system total bandwidth is improved, system average delay is lowered, the matrixes are used for describing a system topology structure, simpleness and convenience are achieved, and computing time during a subnetting process is greatly shortened.

Description

Based on the subnet division methods of 1394b plurality of subnets transmission structure

Technical field

The invention belongs to network information communication technical field, be specifically related to a kind of subnet division methods based on 1394b plurality of subnets transmission structure.

Background technology

IEEE1394 bus was developed and formed in 1987 standard in 1986 by Apple company.Apple company is by its called after Firewire (live wire) at first.Nineteen ninety-five electrical equipment and Electronic Engineering Association (IEEE) be IEEE1394-1995 by FireWire specification; At IEEE in 2000, this agreement is further improved and released IEEE1394a; Within 2002, IEEE has released IEEE1394b based on IEEE1394-1995 and IEEE1394a, and this agreement is supported Optical Fiber Transmission, and bandwidth is supported 3.2Gbps, has greatly improved transmission performance.Because the real-time of IEEE1394b bus is good, transmission rate is high, topology is flexible, expansion is easy and support hot plug, become one of candidate's bussing technique of Aero-Space of future generation, industrial control field.

But 1394b bus is the shared host-host protocol of a kind of traditional bus, and its data are transmitted by shared bus, cannot realize multi-channel data transmission.Although domestic, 1394b bus is possessed some special knowledge, almost do not have about the research of 1394b multiplex techniques.Along with the multiplex demand of Large Volume Data in military field application is more and more higher, the research of the multiplex techniques to 1394b transmission system is imperative.

Different from traditional switching network, in 1394b plurality of subnets switching fabric, form whole data transmission system thereby the port set of all-ones subnet by switch be interconnected, and that node in each subnet is undertaken by annular, chain, shared 1394b fiber buss star-like or tree topology is interconnected, as shown in Figure 1.Each subnet is to adopt traditional 1394b fiber buss interconnect architecture.In the time of subnet internal node mutual data transmission, the separate transfer of data of carrying out of all-ones subnet; In the time having transfer of data between the node of different sub-network, data by switch the source node from a subnet to the destination node in another subnet.According to above-mentioned situation, the transfer of data of plurality of subnets switching fabric is divided into two classes: in subnet, between transmission and subnet, transmit.In subnet, transmission is traditional 1394b fiber buss transmission, and this transmission means has real-time and the certainty of shared bussing technique.Between subnet, transmission is to realize by switch, and its transmission performance is determined by switch.

IEEE has released 1394.1 agreements (1394.1 in 2004 ^tMiEEE Standard for High Performance Serial Bus Bridges, Microprocessor and Microcomputer Standards Committee), it has defined the concept of 1394 bridges, as shown in Figure 2, can be by the network of an intercommunication of many 1394 bus compositions by 1394 bridges, multiple bus can independently carry out transfer of data and not interact, and data also can be transferred in other buses by 1394 bridges, realize the transmission of 1394b network parallel interaction data.

There is defect in above-mentioned 1394 bridge architectures, when packet is transferred in other buses by bridge, it will arrive all buses, take the bandwidth of all buses, comprise the bus of not wishing to obtain this packet, thereby cause whole network efficiency lower.This agreement is not used widely, and does not also support correlative study and the product of this agreement.

Summary of the invention

The problem that while the present invention is directed to 1394b multiplexing, bus bandwidth occupancy is high, network transmission efficiency is lower, based on queueing theory, has proposed a kind of subnet division methods based on 1394b plurality of subnets transmission structure.

Subnet division methods based on 1394b plurality of subnets transmission structure of the present invention, comprises the steps:

Step 1: obtain transmission bandwidth between network node, be expressed as matrix A as follows:

N represents node number, the capable j column element of i a in matrix A _ijrepresentation node i sends the amount of bandwidth of information to node j;

Step 2: establish i subnet P _irepresent network topology matrix X=(P ₁..., P _i..., P _n) _{1 × n}, P _ibe the column vector of n × 1, in the time that certain subnet vector is 0 vector, represent that this subnet does not exist; When r node is at subnet P _iwhen interior, column vector P _ithe value of r row element be 1, otherwise be 0;

Step 3: the affiliated subnet of each node is set, enumerates all-network topological matrix;

Step 4: to each network topology matrix, calculate the time delay in each subnet and between subnet under this network topology, and obtain the average delay of whole network;

Wherein, the time delay in each subnet and between subnet represents by matrix T:

Element T in matrix T _iirepresent subnet P _iinterior time delay, μ is bus average service rate, λ _iifor subnet P _iinterior, subnet P _iand transmission bandwidth total amount between other subnets,

Element T in matrix T _ijrepresent subnet P _iwith subnet P _jbetween time delay, T _ij=T _ii+ T _jj, T _jjrepresent subnet P _jinterior time delay;

The average delay t of whole network is:

$t=<\frac{X^T\mathrm{AX}}{b_{\mathrm{band}}}\&CenterDot;T>$

Wherein, b _bandfor the total bandwidth of network;

Step 5: according to network topology matrix corresponding to whole network average time minimum value, carry out subnet division.

Advantage of the present invention and good effect are:

(1) the present invention has realized 1394b light bus multiplex techniques, is conducive to the application of 1394b light bus;

(2) the inventive method, based on 1394b plurality of subnets transmission structure, has realized 1394b multi-subnet architecture and has divided, and has improved overall system bandwidth, has reduced system average delay;

(3) the inventive method, based on queueing theory, adopts matrix description system topology, simple and easy, has greatly simplified the computing time in subnet partition process.

Brief description of the drawings

Fig. 1 is 1394b plurality of subnets switching fabric schematic diagram;

Fig. 2 is 1394 bridge architecture schematic diagrames;

Fig. 3 is the schematic flow sheet of subnet division methods of the present invention.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

The present invention is directed to the multiplex problem of 1394b, according to 1394b plurality of subnets switching fabric, a kind of simple and easy subnet division methods has been proposed, by transmission bandwidth and the subnet topological structure of node in matrix description 1394b multiloop loop system, set up single window services model, calculate bandwidth and time delay between the each subnet of plurality of subnets system inside, subnet, select the topological structure of time delay minimum, realize the division of 1394b multi-subnet architecture.The present invention not only can obtain the minimum time delay topological structure of system, realizes the subnet of 1394b plurality of subnets transmission structure and divides, and will promote the development of 1394b multiplex techniques.The present invention compares the serial shared bus formula transmission that original 1394b is traditional, has realized the transmission of 1394b multidiameter delay formula, and has improved overall system bandwidth, has reduced communication delay.

The invention provides a kind of subnet division methods based on 1394b plurality of subnets transmission structure, below in conjunction with Fig. 3, specific implementation step is described.

If have n node in the applied 1394b multiplex system of subnet division methods of the present invention, and based on following setting:

(1) the exchanges data amount of each node is known;

Be different from the relatively random service condition of each node in family network, under the actual application background of avionics system, between the node of 1394b multi-subnet architecture, transmission bandwidth is relatively definite, can estimate its size.

(2) in plurality of subnets system, exchanges data meets M/D/1 window scheme;

Based on queueing theory (TQ, the Theory of Queues), obey negative exponent and distribute the interval time of tentation data arrival system, and determine the service time of each packet, transfer of data in subnet in plurality of subnets system, between subnet can be set up to the Mathematical Modeling of M/D/1 window scheme.M/D/1 window scheme is a simple queuing model, and wherein M shows to arrive obedience Markov process, and it is speed generation that determine and to fix that D shows to leave, and 1 shows to only have a station server.

Plurality of subnets division methods of the present invention, the step that the 1394b plurality of subnets topological structure of n node is divided is as follows.

Step 1: obtain transmission bandwidth between network node.

For the network with n node, between all nodes, transmission bandwidth size represents by the matrix A of n × n:

Wherein, the capable j column element of i a in matrix A _ijrepresentation node i sends the amount of bandwidth of information to node j.In network, do not exist node to the situation that self sends data, therefore a _ii=0.

Step 2: establish network topology and be expressed as matrix X, in matrix, each column vector represents a subnet.

For the network with n node, the plurality of subnets topological matrix of network represents by the matrix X of n × n:

X＝(P ₁,…,P _i,…,P _n) _1×n (2)

Wherein, P _irepresenting i subnet, is the column vector of n × 1.When r node is at subnet P _iwhen interior, column vector P _ithe value of r row element be 1, otherwise be 0.In the time that a certain subnet interior element is 0, representing does not have node in this subnet, and this subnet does not exist.

Relatively, in network, each node is for subnet, and the row vector of an available 1 × n represents, when r node is at subnet P _iwhen interior, the vector of corresponding node be (0 ... 0,1,0 ..., 0) _{1 × n}, wherein, r element value is 1, all the other element values are 0.Being added transposition again for the available knot vector that belongs to this subnet of each subnet vector obtains.If subnet P _icomprise k node, vectorial ε wherein _i,krepresent subnet P _iin k node, vectorial P _i=(ε _{i, 1}+ ε _{i, 2}+ ... ε _i,k) ^t.

Step 3: the affiliated subnet of each node is set, enumerates all-network topological matrix.

By subnet under each node is set, can obtain different network topologies.For the network of n node, the total n of network topology matrix that may exist ⁿindividual, consider network symmetry, can calculate half quantity (n below ⁿ/ 2) network topology matrix.

Step 4: determine the transmission bandwidth in each subnet and between each subnet under current network topology, and obtain the average delay of whole network.

In network, between arbitrary node, amount of bandwidth can be expressed as:

$\left.\begin{array}{c}{a}_{\mathrm{ij}}={\mathrm{\&epsiv;}}_i^{T}A{\mathrm{\&epsiv;}}_j\\ a_{\mathrm{ii}}={\mathrm{\&epsiv;}}_i^{T}A{\mathrm{\&epsiv;}}_i\end{array}\right\}---\left(3\right)$

ε _i, ε _jrepresent respectively the vector of node i, node j.

For the plurality of subnets system with n node, in each subnet, between subnet, transmission bandwidth size represents by the matrix B of n × n:

Wherein, b _iirepresent the inner each node transmission bandwidth total amount of subnet i, if subnet inside only has 1 node or there is no node, b _iibe 0; b _ijrepresent the bandwidth sum to each node transmission in subnet j of each node in subnet i.

Through checking, transmission bandwidth size matrix A and its topological structure matrix X between node, and between each subnet, subnet, the relation of transmission bandwidth matrix B meets following formula:

B＝X ^TAX (5)

Proof procedure:

If subnet P _iin have k node: i ₁, i ₂..., i _k, subnet P _iin data transfer bandwidth be data transfer bandwidth sum between any two nodes wherein:

$b_{\mathrm{ii}}=a_{i_1{i}_1}+a_{i_1{i}_2}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_1{i}_k}+a_{i_2{i}_1}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_2{i}_k}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_k{i}_1}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_k{i}_k}---\left(6\right)$

Bring (3) formula into and can obtain subnet P _iinterior data transfer bandwidth b _iifor:

$\begin{array}{c}{b}_{\mathrm{ii}}={\mathrm{\&epsiv;}}_{i_1}^{T}A{\mathrm{\&epsiv;}}_{i_1}+{\mathrm{\&epsiv;}}_{i_1}^T{\mathrm{A\&epsiv;}}_{i_2}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&epsiv;}}_{i_1}^{T}A{\mathrm{\&epsiv;}}_{i_k}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&epsiv;}}_{i_k}^T{\mathrm{A\&epsiv;}}_{i_1}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&epsiv;}}_{i_k}^T{\mathrm{A\&epsiv;}}_{i_k}\\ =({\mathrm{\&epsiv;}}_{i1}^T+{\mathrm{\&epsiv;}}_{i2}^T+\&CenterDot;\&CenterDot;\&CenterDot;{\mathrm{\&epsiv;}}_{\mathrm{ik}}^T)A({\mathrm{\&epsiv;}}_{i_1}+{\mathrm{\&epsiv;}}_{i_2}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&epsiv;}}_{i_k})=P_i^{T}A{P}_i\end{array}---\left(7\right)$

In like manner, can obtain the data transfer bandwidth between subnet.If subnet P _iin have m node: i ₁, i ₂..., i _m, subnet P _jin have h node: j ₁, j ₂..., j _h, subnet P _iwith subnet P _jbetween data transfer bandwidth b _iifor:

$\begin{array}{c}{b}_{\mathrm{ij}}=a_{i_1{j}_1}+a_{i_1{j}_2}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_1{j}_h}+a_{i_2{j}_1}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_1{j}_h}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_m{j}_1}+\&CenterDot;\&CenterDot;\&CenterDot;+a_{i_m{j}_h}\\ =({\mathrm{\&epsiv;}}_{i1}^T+{\mathrm{\&epsiv;}}_{i2}^T+\&CenterDot;\&CenterDot;\&CenterDot;{\mathrm{\&epsiv;}}_{\mathrm{in}}^T)A({\mathrm{\&epsiv;}}_{j_1}+{\mathrm{\&epsiv;}}_{j_2}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&epsiv;}}_{j_n})=P_i^T{\mathrm{AP}}_j\end{array}---\left(8\right)$

(7) and (8) two formulas are brought in formula (4),

Be B=X ^taX.

Therefore, the topological matrix X of matrix A and plurality of subnets system by transmission bandwidth size between node, can obtain transmission bandwidth matrix B in the subnet of 1394b plurality of subnets system, between subnet.

For transmission bandwidth size matrix B in each subnet of n × n, between subnet, in corresponding subnet, between subnet, transmission delay represents by the matrix T of n × n:

Wherein, T _iirepresent subnet P _iinterior time delay.When the interior transfer of data of subnet, node is in a 1394b bus, calculate the average latency and need to consider that all and this bus sends the node of data (comprising in subnet and between subnet), calculates average queue waiting time formula according to M/D/1 window scheme wherein λ is data packet number, and μ is bus average service rate.For subnet P _i, λ is subnet P _iinterior, subnet P _iand transmission bandwidth total amount λ between other subnets _ii:

$\begin{array}{c}{\mathrm{\&lambda;}}_{\mathrm{ii}}=b_{i1}+b_{i2}+\&CenterDot;\&CenterDot;\&CenterDot;+b_{\mathrm{in}}+b_{1i}+b_{2i}+\&CenterDot;\&CenterDot;\&CenterDot;b_{\mathrm{ni}}-b_{\mathrm{ii}}\\ =P_i^T\mathrm{AX}+X^T{\mathrm{AP}}_i-P_i^T{\mathrm{AP}}_i\end{array}---\left(11\right)$

Time delay can obtain.

In like manner, T _ijfor subnet P _iwith subnet P _jbetween time delay.Data are by subnet P _imiddle node transfers to switch and is transferred to subnet P again _jmiddle node, because the time delay of switch was considered in calculating subnet internal delay time, so T _ij=T _ii+ T _jj.

Finally, the average delay t that obtains whole network is

$t=<\frac{B}{b_{\mathrm{band}}}\&CenterDot;T>=<\frac{X^T\mathrm{AX}}{b_{\mathrm{band}}}\&CenterDot;T>---\left(12\right)$

Wherein, b _bandfor the total bandwidth of network.Formula (12) represents that the bandwidth ratio time delay product corresponding with it that each subnet is taken is added, and calculates the average delay of plurality of subnets system.

Step 5: according to network topology matrix corresponding to whole network average time minimum value, carry out subnet division.

The topological matrix of whole network average time minimum is t _mintime corresponding X _minmatrix,

$t_{\min }=<\frac{X_{\min }^T{\mathrm{AX}}_{\min }}{b_{\mathrm{band}}}\&CenterDot;T>---\left(13\right)$

Claims (1)

1. the subnet division methods based on 1394b plurality of subnets transmission structure, is characterized in that, comprises the steps:

Step 1: obtain transmission bandwidth between network node, be expressed as matrix A as follows:

N represents node number, the capable j column element of i a in matrix A _ijrepresentation node i sends the amount of bandwidth of information to node j;

Step 2: establish i subnet P _irepresent network topology matrix X=(P ₁..., P _i..., P _n) _{1 × n}, P _ibe the column vector of n × 1, in the time that certain subnet vector is 0 vector, represent that this subnet does not exist; When r node is at subnet P _iwhen interior, column vector P _ithe value of r row element be 1, otherwise be 0;

Step 3: the affiliated subnet of each node is set, enumerates all-network topological matrix;

Step 4: to each network topology matrix, calculate the time delay in each subnet and between subnet under this network topology, and obtain the average delay of whole network;

Wherein, the time delay in each subnet and between subnet represents by matrix T:

Element T in matrix T _iirepresent subnet P _iinterior time delay, μ is bus average service rate, λ _iifor subnet P _iinterior, subnet P _iand transmission bandwidth total amount between other subnets,

Element T in matrix T _ijrepresent subnet P _iwith subnet P _jbetween time delay, T _ij=T _ii+ T _jj, T _jjrepresent subnet P _jinterior time delay;

The average delay t of whole network is:

$t=<\frac{X^T\mathrm{AX}}{b_{\mathrm{band}}}\&CenterDot;T>$

Wherein, b _bandfor the total bandwidth of network;

Step 5: according to network topology matrix corresponding to whole network average time minimum value, carry out subnet division.