CN100405782C - Optical Burst Switching Routing Method Based on Resource Prediction - Google Patents

Abstract

The invention discloses a resource prediction-based optical burst switching route selection method, which includes reading three-state information of a first-level node in a database: transmission distance L ₁ , refresh cycle T ₁ , and wavelength busyness P ₁ and wavelength occupancy information ψ ₁ ; read the corresponding three-state information of the secondary node; respectively substitute each parameter into the model based on linear or nonlinear discrete information prediction, and the calculated result is the reliability of the state information of the corresponding node The correction factor ε (0<ε<1), and together with the original wavelength information ψ (ψ=0, 1) of the node, calculates the instant resource information of the node. If ψ is 0, it is corrected to 0+ε, and when ψ is 1, it is corrected to 1-ε; the array A is composed of the real-time resource information of all nodes in the first-level nodes, and the array A is also formed by the information of all nodes in the second-level nodes. Each real-time resource information is an object to form an array B, and Q=A×B' is calculated; the two nodes corresponding to the largest Q value are selected as the maximum reliability path for burst packet transmission. The invention will greatly reduce the blocking probability and packet loss rate of burst packets, and greatly improve the utilization rate of bandwidth.

Description

Optical Burst Switching Routing Method Based on Resource Prediction

technical field

The invention relates to a resource prediction-based optical burst switching routing selection method.

Background technique

With the rapid growth of IP services around the world, the demand for transmission network bandwidth and switching system capacity is increasing at an unprecedented rate. At present, with the maturity of Dense Wavelength Division Multiplexing (DWDM), it is possible to make full use of the massive bandwidth of optical fibers (each optical fiber can exceed 10Tbps). As a result, the switching equipment part becomes the bottleneck of the communication network. At present, the speed of commercialized IP routers can reach 10-120Gbps, but there is still a big gap compared with the transmission capacity of DWDM. Realizing all-optical switching is a feasible solution to this bottleneck. According to the existing communication mode in the communication network, people put forward three schemes on how to realize switching and transmission in the WDM optical network:

1) Optical path switching is the optical wavelength routing switching solution (OCS). Wavelength routing belongs to coarse-grained switching, and its route selection and link establishment time are relatively long. flexibility.

2) Optical Packet Switching (OPS), but due to the lack of high-speed optical logic devices, optical buffer memory, etc., it is still in the research stage and has not been put into practical applications.

3) Optical Burst Switching (OBS), which is a comprehensive and compromise solution, which combines the advantages of OPS and OCS and overcomes the shortcomings of both. Relatively speaking, it is more flexible than wavelength routing and bandwidth utilization It is a switching technology that is most likely to be practical at present.

In optical burst packet switching, at the source node, the sending process of a burst packet is usually as follows: grouping packets, selecting routes, sending signaling, and sending data. At present, the burst packet length is generally on the order of ms, and the routing decision time is generally on the order of ns. As shown in Figure 1, assuming that at time _t1 , the source node receives status update messages from other nodes, until the next time it receives The time for other node status update messages is generally in the order of s, which is much longer than the routing decision time, and the time interval for updating the routing table is uncertain, so it brings great uncertainty to routing selection sex factor. When the source node A selects a route at time _t2 , if the status information of each node has not been refreshed at this time, the usual practice is to consider that the idle probability of the channel that is idle at time _t1 is 1 when the burst packet arrives at the channel, that is, The credibility of the selected path with an idle probability of 1 will decrease as the time distance from the last update increases, as shown in Figure 2. The abscissa in the figure is time, and the ordinate is the credibility of the information in the routing information table. The origin is the refresh time of a certain routing information table. This figure shows the routing time interval from this refresh to the next refresh. The reliability of table information decreases with time. The method of the invention is to add a prediction mechanism of route selection on the basis of common routing algorithm. In this way, the entire network is regarded as a quasi-static network. However, other nodes may also be selecting routes between two state refreshes. If other nodes also select the channel selected by source node A, then when two burst packets arrive at the same time When the channel is blocked, a burst of data packets will be blocked. This route selection method will cause a relatively large blocking rate and packet loss rate in an actual network, especially when the network is relatively large and in an optical burst switching network using the Just in Time protocol.

Contents of the invention

Aiming at the problems and deficiencies in the burst packet routing in the existing optical burst switching, the purpose of the present invention is to provide a resource prediction-based optical packet that can instantly calculate the occupancy of nodes and select non-busy channels in real time. Burst switching routing method.

The present invention is achieved in this way: a resource prediction based optical burst switching routing method, comprising the following steps,

1) When the source node is grouping optical burst packets, it reads the three-state information of the first-level node in the corresponding database: transmission distance L ₁ , refresh cycle T ₁ , wavelength busyness P ₁ and wavelength occupancy information ψ ₁ ; Similarly, read the three-state information of the secondary node: transmission distance L ₂ , refresh cycle T ₂ , wavelength busyness P ₂ and ψ ₂ ;

2) Substitute the parameters in step 1) into models based on linear or nonlinear discrete information prediction, such as Wiener filtering method, Kalman filtering method, and continuous mathematical functions such as joint Gaussian membership function, bell membership Degree function, sigmoid function type membership function, trapezoidal membership function or triangular membership function medium information prediction mechanism, the calculated result is the reliability correction factor ε(0<ε<1) of the state information of the corresponding node, And calculate the real-time resource information of the node together with the original wavelength information ψ(ψ=0, 1) of the node, specifically, if ψ is 0, it is corrected to 0+ε, and when ψ is 1, it is corrected to 1-ε;

3) form an array A with each instant resource information of all nodes in the first-level nodes as objects, and form an array B with each instant resource information of all nodes in the second-level nodes as objects, and ask Q=A×B′; A set of revised confidence state information is obtained.

4) Select the two nodes corresponding to the largest maximum credibility Q value as the maximum credibility path for burst packet transmission.

Further, the wavelength occupancy information sent by the node to the state resource database of the whole network has only two states, that is, "0" and "1" states, wherein "0" state means occupied, and the availability probability is 0; "1" means Idle, availability probability is 1.

When the present invention selects a route, it is considered that the idle channel at a certain moment _t1 starts from the moment _t1 , and its idle probability is not necessarily 1, which changes with time, usually the farther away from the moment _t1 , the idler The confidence of the probability is smaller. Routing should select the channel with the highest idle probability (maximum product) from the source node to the destination node at the current moment to send burst packet data. The invention will greatly reduce the blocking probability and packet loss rate of burst packets, and greatly improve the utilization rate of bandwidth.

Description of drawings

The present invention will be described in detail below in conjunction with the accompanying drawings.

Figure 1 is a schematic diagram of the refresh time of status information;

Fig. 2 is a schematic diagram of state information credibility;

Fig. 3 is a schematic diagram of routing;

Fig. 4 is a flowchart of the wavelength prediction mechanism of the present invention;

Fig. 5 is an example of the whole network state information database of the present invention.

Detailed ways

As shown in Figure 2, when the present invention selects a route at _t2 , it is considered that the idle channel whose state is refreshed at a certain moment _t1 starts from time _t1 , and its idle probability is not necessarily 1, and its idle probability can be determined The reliability varies with time from time t ₁ to the next refresh time t ₃ , usually the farther away from time t ₁ , the smaller the reliability of the idle probability is 1. After the predictive mechanism is added during route selection, the blocking probability and packet loss rate of burst packets will be reduced. Compared with the original data communication flow, the improved flow of the source node sending a burst packet in the present invention becomes: grouping→forecasting→route selection→sending signaling→sending a burst data packet.

The present invention can use the discrete information prediction method and/or the mathematical function prediction method to predict the resource state of the node occupancy for the resource state information. For the discrete information prediction mechanism, the present invention adopts linear and nonlinear prediction models such as Wiener filter and Kalman filter. As for the mathematical prediction function, the fuzzy membership function is selected to predict the resource state. For a fuzzy number, the degree of membership indicates how fuzzy the value is. The present invention provides factors affecting the degree of unreliability of resource status information, which are summarized into the following aspects:

1. The ability of nodes to process information. If the processing capacity of the node is low, the processing delay may not be able to keep up with the change of the dynamic call.

2. The distance of status information transmission. The farther the node is from the source node, the higher the unreliability of the information.

3. The rate at which node status information is refreshed. The shorter the refresh cycle, the faster the rate and the higher the reliability of the information.

4. The historical information of each node in the edge node network-wide state information database.

In the present invention, the wavelength occupancy information sent by each core node to the state resource database of the whole network has only two states, that is, two states of "0" and "1". The present invention defines that "0" means occupied, and the available probability is 0; "1" means idle, and the available probability is 1. Therefore, according to the "0" and "1" state distribution of the wavelength resource history information of each node in the resource database of the whole network, the busyness of the corresponding wavelength and the probability that it will be occupied can be judged.

As shown in Figure 3, when the burst packet arrives from the source node, the last refreshed state information database of the entire network shows that only A1 is available for the first-level node, and only B2 is available for the second-level node, so the route selection is indicated by the arrow in the figure In fact, from the last time the state information database of the whole network was updated to the present moment, the state of the network has been changing dynamically during this period, so it cannot be considered that the path selected at this time is idle with probability 1 , while the other 8 paths are occupied with probability 1.

As shown in Figure 4, the route selection method of the present invention is to read the three-state information of the first-level node in the corresponding database: the transmission distance L ₁ , the refresh cycle T ₁ , the wavelength busyness P ₁ and the wavelength occupancy information ψ ₁ ; similarly, read the three-state information of the secondary node: transmission distance L ₂ , refresh period T ₂ , wavelength busyness P ₂ and ψ ₂ ; respectively substitute the above parameters into linear and nonlinear discrete information prediction Models such as Wiener filter, Kalman filter, etc., or mathematical functions such as joint Gaussian membership function, bell-shaped membership function, sigmoid function-type membership function, trapezoidal membership function or triangular membership function and other prediction functions, The calculated result is the correction factor ε(0<ε<1) of the state information of the corresponding node, and together with the original wavelength information ψ(ψ=0,1) of the node, the real-time resource information of the node is calculated, specifically as , if ψ is 0, it is corrected to 0+ε, and when ψ is 1, it is corrected to 1-ε. Figure 5 shows an example of the state table in the state information database of the whole network according to the flow of the prediction mechanism. It can be clearly seen from the chart that the refreshed state information of node N has a long transmission distance and a busy wavelength, so the information returned to the state database of the whole network has a high degree of ambiguity and low reliability. Therefore, the revised state is consistent with the original information On the contrary; use the real-time resource information of all 3 nodes in the first-level node as the object to establish array A, $A=\left[\begin{array}{c}1-{\mathrm{\&epsiv;}}_{A1}\\ 0+{\mathrm{\&epsiv;}}_{A2}\\ 0+{\mathrm{\&epsiv;}}_{A3}\end{array}\right];$ Similarly, take the real-time resource information of all three nodes in the second-level node as the object

build array B, $B=\left[\begin{array}{c}0+{\mathrm{\&epsiv;}}_{\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="C20041000973900082.png,C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout>}1}\\ 1-{\mathrm{\&epsiv;}}_{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout>}2}\\ 0+{\mathrm{\&epsiv;}}_{\mathrm{<figure-callout\; id="3"\; label="B"\; filenames="C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout>}3}\end{array}\right];$ Multiplying matrix A by the transpose of B gives,

$\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="3"\; label="B"\; filenames="C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout></span>}\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="1"\; label="B"\; filenames="C20041000973900082.png,C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout></span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>& {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="3"\; label="B"\; filenames="C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout></span>}\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="1"\; label="B"\; filenames="C20041000973900082.png,C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout></span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>& {\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="3"\; label="B"\; filenames="C20041000973900101.png"\; state="\{\{state\}\}">B</figure-callout></span>}\mathrm{<span\; class="notranslate">\; 3</span>}}\end{array}\right]<span\; class="notranslate">\; .</span>$

By comparing the size of each item Q _ij in the matrix Q, select the largest Q _ij , and its corresponding i and j are the paths to be selected by the first-level node and the second-level node respectively. In this way, through a prediction mechanism, the blocking rate and packet loss rate of data burst packets can be reduced, and the communication efficiency of the network can be improved.

Claims (2)

1. the light burst-switched route selection method based on resources may further comprise the steps,

When 1) the source node burst packets arrives, in associated databases, read the three condition information of one-level node: the distance L of transmission ₁, the period T that refreshes ₁, wavelength busy extent P ₁With wavelength occupied information ψ ₁Similarly, read the three condition information of two-level node: the distance L of transmission ₂, the period T that refreshes ₂, wavelength busy extent P ₂With wavelength occupied information ψ ₂

2) each parameter in the step 1) is updated to respectively based on linear and nonlinear prediction model, or mathematical function is as associating Gaussian membership function, the bell membership function, sigmoid function type membership function, in trapezoidal membership function or the Triangleshape grade of membership function anticipation function, the result who calculates is the modifying factor ε of the state information of respective nodes, and calculate the instant resource information of this node with the former wavelength information ψ of this node, 0＜ε＜1 wherein, ψ=0,1, be specially, if ψ was modified to 0+ ε at 0 o'clock, ψ was modified to 1-ε at 1 o'clock;

3) each the instant resource information with all nodes in the one-level node is that object is formed array A, and each the instant resource information with all nodes in the two-level node is that object is formed array B equally, and asks Q=A * B ', and wherein B ' is the transposition of B;

4) select the path of two corresponding nodes of a maximum Q value as the burst packets transmission.

2. the light burst-switched route selection method based on resources as claimed in claim 1 is characterized in that, the wavelength occupied information that described node sends to the whole network state resource database is two states only, i.e. " 0 " one state, wherein, " 0 " attitude is represented to take, and can utilize probability is 0; " 1 " expression is idle, and can utilize probability is 1.

Images