CN1960283A - Performance evaluation method for communication mode in sub layer to be adopted to transmission of expandable route system - Google Patents

Abstract

The traditional communication mode between plane nodes controlled by expandable router has the expandable bottle neck and needs to be improved. The invention provides an estimation method for the improved mode, which introduces the concepts of the traffic between planes and the increment factor of traffic between planes; and the distributing rate, dispersion rate and traffic rate of control plane task are used as the characteristic parameters for estimating the performance. The estimation method based on the mode can estimate the communication performance and newly-received load.

Description

Expandable route system transmits the method for evaluating performance of adaptive sublayer communication pattern

Technical field

The method of evaluating performance that expandable route system transmits adaptive sublayer communication pattern belongs to extendible router software architecture technical field.

Background technology

Router is a server equipment, but the router architecture with extended attribute has become one of the important research direction of router of future generation and difficult point problem.Notion from the general function structure, router can be divided into two main function planes: one is to concentrate on flow to transmit the datum plane of handling, another is to finish control and the mutual control plane of route, datum plane is the main information source of control plane to the information flow of control plane, directly influences the operation conditions of software architecture.

Because the simplicity of datum plane, in traditional router software architecture, from the control information on bottom data plane, for example the mutual message of route etc. all is a raw information, and information does not process before handling arriving control plane.When if the node of a plurality of control planes is arranged, data are not done each node that is submitted to control plane with distinguishing, see Fig. 1.The node of each control plane is only selected and this node Useful Information is submitted to each corresponding protocol is handled after receiving original control information.This communication pattern (being the inundation pattern) can operate as normal under the single or number situation seldom at the control plane node, but along with expandable route retrains the nothing of control plane interstitial content in principle, this pattern can cause by the Control Node number and increase and the control flows communication performance bottleneck that brings and inner communication bandwidth waste, has influenced the extended capability of software architecture greatly.We improve this pattern, see Fig. 2, by between datum plane and control plane, having inserted one " transmitting adaptive sublayer ", repeat flow in interplanar transmission to reduce as far as possible, communication can be expanded the purpose of bottleneck between the elimination face of reaching, for distinguishing traditional inundation pattern, we are called the communication pattern that transmits adaptive sublayer with this communication pattern, and structure chart is seen Fig. 3.

The objective of the invention is to propose a kind of evaluation method to this new traffic pattern communication performance.Because the purpose of improving communication pattern is to improve the efficient communication rate of control plane, but because the efficient communication rate itself is an absolute value, can not reflect the improvement of performance.We propose traditional pattern and the delivery flow rate of the pattern after improving compares, and then the performance of improved mode is estimated with the ability of holding load, and provide corresponding task optimization allocative decision according to analysis.

Summary of the invention

The object of the present invention is to provide a kind of expandable route system to transmit the method for evaluating performance of adaptive sublayer communication pattern.

The invention is characterized in, contain following steps successively:

Step 1. initialization;

At first introduce the following definitions for evaluation method:

Flow between face: the information flow total amount that datum plane is flowed through to control plane is called flow between face;

Flow extender index between face: under fixing task and flow rate mode, the ratio of flow between the face of flow and a contrastive pattern between inundation pattern face is called between this contrastive pattern's face and flows extender index, remember and make II _c, between face the stream extender index has been expressed between contrastive pattern's face flow to " the saving degree " of bandwidth, and the inverse of stream extender index has been expressed a contrastive pattern and will have been reached flow between the face of the inundation pattern that not have optimization between face, can also additionally hold the ability of load;

Each task in the step 2. pair system is set its configuration parameter respectively:

Step 2.1. is provided with the quantity that t is a task in the system;

Step 2.2. is provided with total number that m is the node that distributes of task on control plane;

The flow that step 2.3. is provided with each node in each task respectively accounts for the ratio of task total flow, D=(d ₁..., d _t), 0≤d wherein _i≤ 1, be called the distributive law of task i, for the sake of simplicity, (flow that each node distributes among the i＜=t) is identical to suppose any task i;

Step 2.4. is provided with the number of the node that each task may be distributed to, E=(e respectively ₁..., e _t), 1≤e wherein _i≤ m is called the dispersion number of task i;

The flow that step 2.5. is provided with each task respectively accounts for the ratio of total flow, C=(c ₁..., c _t), 0≤c wherein _i≤ 1, be called the rate of discharge of task i;

The flow of task i in the step 3. calculation procedure 2 described systems:

I(i)＝c _i+d _i(e _i-1)c _i＝(1+d _i(e _i-1))c _i；

The calculating of flow during all t task in the step 4. pair system carry out step 3, and flow between the face of the system that obtains that sues for peace:

$I=\underset{i=1}{\overset{t}{\mathrm{\Σ}}}I\left(i\right)=\underset{i=1}{\overset{t}{\mathrm{\Σ}}}(1+d_i(e_i-1))c_i;$

Step 5. calculation procedure 2 described systems flow between the face under the inundation pattern:

Make the general formula d in the step 4 _i=1, e _i=m then obtains flow I=mc between face under the inundation pattern _i

Step 6. is calculated and is flowed extender index between the face that transmits adaptive sublayer pattern:

${\mathrm{II}}_c={\left(\frac{{\mathrm{\Σ}}_{i=1}^t(1+d_i(e_i-1))c_i}{{\mathrm{\Σ}}_{i=1}^t{\mathrm{mc}}_i}\right)}^{-1};$

Have according to definition ${\mathrm{\Σ}}_{i=1}^t{c}_i=1,$ Then

${\mathrm{II}}_c={\left(\underset{i=1}{\overset{\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A20061011387200102.png,A20061011387200111.png"\; state="\{\{state\}\}">t</figure-callout>}}{\mathrm{\&Sigma;}}}\frac{1+d_i(e_i-1)}{m}{c}_i\right)}^{-1};$

Step 7. is estimated system according to stream increment factor between face:

Step 7.1. definition: order ${\mathrm{\&beta;}}_i=\frac{1+d_i(e_i-1)}{m},$ The flow distribution feature that it has reflected a task is to the capability of influence of stream extender index between the face of system, is called between the face of task to flow increment factor;

Step 7.2. is in conjunction with II _cAnd β _iFormula as can be known: ${\mathrm{II}}_c={\left(\underset{i=1}{\overset{t}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_i{c}_i\right)}^{-1},$ The product of the rate of discharge of stream increment factor and task is exactly the contribution of stream extender index inverse between a task opposite between face; The inverse of stream extender index just equals the inner product of stream increment factor and rate of discharge between the face of task between face, in vector space, the increment factor that do not flow between coplanar of all tasks of system makes up the bandwidth conservation degree that the projection of its rate of discharge vector has just been characterized this contrastive pattern, also just characterized admittance ability to new adding load, thereby can determine to increase the corresponding regular load of each task, i.e. the load of describing by the problem quantity that task solved;

The influence of stream extender index inverse and the optimization of Task Distribution pattern (being the allocation situation of t task on m node) between the distributive law of step 8. task and dispersion number opposite:

The influence of stream extender index inverse between the distributive law of step 8.1. task and dispersion number opposite:

According to the formula of flow between face can obtain between face stream extender index inverse to any task i about the partial derivative of distributive law with about disperseing the difference of number:

$\frac{\&PartialD;{\left({\mathrm{II}}_c\right)}^{-1}}{\&PartialD;d_i}=\frac{(e_i-1)c_i}{m};$

$\frac{\mathrm{\&Delta;}{\left({\mathrm{II}}_c\right)}^{-1}}{\mathrm{\&Delta;}{e}_i}=\frac{d_i{c}_i}{m};$

The allocation model optimization of step 8.2. task:

For rate of discharge: be one of intrinsic characteristic of system, also with the load environmental correclation, and with adopted which type of Task Distribution pattern irrelevant;

For distributive law with disperse number: stream extender index inverse is described about their rate of change such as step 8.1 between face; The parameter that can select the speed maximum thus is as definite direction of improving the task distribution pattern.

The present invention can estimate the communication performance of transmitting adaptive sublayer communication pattern, based on traditional inundation communication pattern, define the foundation that the stream extender index is estimated as model performance between flow between face and relevant face, the distributive law by the control plane task, disperseed number and three characteristic parameters of rate of discharge that the performance of model is analyzed; And the distribution to task has proposed the reference scheme of optimizing.

Description of drawings

Fig. 1. the communication structure between datum plane and control plane;

Fig. 2. the logic communication structure after the improvement;

Fig. 3. stream increment schematic diagram between face;

Fig. 4. the variation of stream increment factor between face;

Fig. 5. experimental result;

Fig. 6. the variation of stream increment factor between face;

Fig. 7. experimental result.

Embodiment

Situation with two tasks is an example, and as Fig. 5, horizontal ordinate is represented respectively between the face of two tasks and flowed increment factor.(O C) is the rate of discharge vector of two tasks.According to the definition of flowing increment factor between face as can be known:

0≤d _i≤1；

1≤e _i≤m；

∴ $\frac{1}{m}\&le;{\mathrm{\&beta;}}_i\&le;1,$ i∈[1，…，m]；

Therefore, figure mid point A is the lower bound towards stream increment factor span, and some B is the upper bound towards stream increment factor span.Dash area is exactly the scope that stream increment factor vector can value between face.Under the notion of only considering relative size, dash area arbitrfary point (β ₁, β ₂) inverse of stream increment is exactly that it is in vector (O, C) projected length between institute's corresponding surface.(β ₁, β ₂) be projected as D ', A, the projection of B is respectively A ', B '.Between face stream extender index span reciprocal be exactly (A, B).The stream increment factor makes up the bandwidth conservation degree that corresponding projection has just characterized the contrastive pattern between two task faces.Flow increment factor because the inundation pattern is put between pairing of B exactly, therefore from figure, it can also be seen that among the contrastive pattern, the admittance ability of new adding load.

The projection of B just and (β ₁, β ₂) length ratio of projection

As seen from Figure 6, dark shaded areas is the optimization zone of stream increment factor between face.As point (β ₁, β ₂) regional when mobile to this, stream extender index inverse can further reduce between the face of system, and system can hold more load.But, the Changing Pattern of stream increment factor and inequality, i.e. l among the figure between the face of task ₁And l ₂The face with task of moving between stream feature and distribution characteristics relevant, this relation is definite jointly by the rate of discharge of stream increment factor and task between face.

Can calculate by top analysis that system can increase flow load between much newly after introducing improved mode, the also just corresponding corresponding regular load that can increase each task.Below problem be exactly how under new model, come performance is optimized according to discharge characteristic between the face of each task.In Fig. 6, just to determine l ₁, l ₂Rate travel under different parameters is selected the optimized distribution reference scheme of the parameter of speed maximum as task.

The BGP route iterative model based on tree is adopted in experiment, gets 4 node iteration trees of k=2, except the simulation task of BGP, also has a management role.Experiment parameter be set as follows table:

Table 2 experiment parameter is set

Parameter item

The BGP task

Management role

Distributive law (d i) dispersion number (e i) rate of discharge (c i)

5％-100％ 3 0.9

5％ 4 0.1

Wherein the BGP task distribution is on three leaf nodes of iteration tree, and management role is distributed on all nodes.This distributed model defines the allocation model of task, thereby the dispersion number of task is fixed.Our experiment changes the flow distribution rate of BGP task, measures flow between the face of the adaptive sublayer of the transmission communication pattern under inundation pattern and the different parameters respectively, thereby obtains flowing extender index between face under this contrastive pattern.

In the experiment of reality, BGP only arrives time leaf node from the control flows data of datum plane, so the task distribution joint

Count and do not get m=4, but order to two tasks given different task allocation nodes count m _BGP=3 and m _MGMT=4, this is more approaching with actual route system.Rate of discharge from 5% to 100% between the face of BGP task, and incremental steps gets 5%.Result of experiment as shown in Figure 7.

The stream extender index goes out with asterisk in the drawings between the face that mensuration obtains under different parameters.As we can see from the figure, along the direction that distributions rate between the face of BGP task reduces, the stream extender index increases gradually between face, this means and can hold flow load between bigger face.When rate of discharge has only 5% between the face of BGP task, can hold and be equivalent to the about 4.5 times load of the inundation pattern limit.This only sends to corresponding non-leaf node with the route updating packet of BGP, and other practical situation for the control messages of broadcasting are suitable.Another aspect, stream extender index theoretical value between with dashed lines has provided below corresponding parameter among the figure.The result of measuring and theoretical value are identical substantially on trend, numerically are slightly less than theoretical value, and this is owing to exist in experimentation between extra face due to the flow expense.

This shows that the present invention has reached intended purposes.

Claims (1)

1. expandable route system transmits the method for evaluating performance of adaptive sublayer communication pattern, it is characterized in that described method contains following steps successively:

Step 1. initialization:

For each task in the described system is set following configuration parameter respectively:

T is the quantity of task in the system;

M, total number of the node that on control plane, distributes for task;

(under all identical hypothesis prerequisite of the flow that each node distributes among the i＜=t), the flow that each node in each task is set respectively accounts for the ratio of task total flow, represents D=(d with vectorial D at any task i ₁..., d _t), 0≤d wherein _i≤ 1, be called the distributive law of task i;

The number of the node that each task may be distributed to is set respectively, represents E=(e with vectorial E ₁..., e _t), 1≤e wherein _i≤ m is called the dispersion number of task i;

The flow that each task is set respectively accounts for the ratio of task total flow, represents C=(c with vectorial C ₁..., c _t), 0≤c wherein _i≤ 1, be called the rate of discharge of task i;

Step 2. is calculated the flow I (i) of the task i under the communication pattern of the adaptive sublayer of transmission of described system:

I(i)＝c _i+d _i(e _i-1)c _i＝(1+d _i(e _i-1))c _i；

All t task in the step 3. pair described system is carried out the calculating described in step 2, and the summation obtain the information flow total amount that datum plane is flowed through to control plane, be defined as flow I between face:

$I=\underset{i=1}{\overset{t}{\mathrm{\&Sigma;}}}\mathrm{II}\left(i\right)=\underset{i=1}{\overset{t}{\mathrm{\&Sigma;}}}(1+d_i(e_i-1))c_i;$

Step 4. is calculated described system flow, that is: d between the face under the inundation pattern _i=1, e _i=m, I=mc at this moment _i

The ratio of flow is defined as between the face that transmits adaptive sublayer communication pattern and flows extender index between the face under two kinds of communication patterns of step 5. calculating inundation pattern and the adaptive sublayer of transmission, is designated as II _c:

${\mathrm{II}}_c={\left(\frac{{\mathrm{\&Sigma;}}_{i=1}^t(1+d_i(e_i-1))c_i}{{\mathrm{\&Sigma;}}_{i=1}^t{\mathrm{mc}}_i}\right)}^{-1};$

Have according to definition ${\mathrm{\&Sigma;}}_{i=1}^t{c}_i=1,$ Then

${\mathrm{II}}_c={\left(\underset{i=1}{\overset{t}{\mathrm{\&Sigma;}}}\frac{1+d_i(e_i-1)}{m}{c}_i\right)}^{-1};$

Step 6. is estimated system according to the following steps according to stream increment factor between the face of task, stream increment factor β between described _iExpression, it has reflected the capability of influence of the flow distribution feature of a task to stream extender index between the face of system:

${\mathrm{\&beta;}}_i=\frac{1+d_i(e_i-1)}{m};$

Obtain: ${\mathrm{II}}_c={\left(\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_i{c}_i\right)}^{-1},$ The bandwidth conservation degree that has reflected the adaptive sublayer of described transmission communication pattern;

Step 7. is according to the distributive law d of following task _iWith dispersion number e _iThe influence of stream extender index between the opposite is optimized the allocation situation of t task on m node:

Stream extender index inverse is to the partial derivative of any task i about distributive law between step 7.1. calculating face, and about disperseing the difference of number:

$\frac{\&PartialD;{\left({\mathrm{II}}_c\right)}^{-1}}{\&PartialD;d_i}=\frac{(e_i-1)c_i}{m};$

$\frac{\mathrm{\&Delta;}{\left({\mathrm{II}}_c\right)}^{-1}}{\mathrm{\&Delta;}{e}_i}=\frac{d_i{c}_i}{m};$

Step 7.2. is according to the result of calculation of step 7.1, in system to the suitable d of each task choosing _iWith e _i, make $\frac{\&PartialD;{\left({\mathrm{II}}_c\right)}^{-1}}{\&PartialD;d_i},\frac{\mathrm{\&Delta;}{\left({\mathrm{II}}_c\right)}^{-1}}{\mathrm{\&Delta;}{e}_i}$ Maximum.

Images