JP4831542B2 - Wavelength allocation optimization calculation method and collision avoidance method by wavelength allocation control in synchronous optical packet switching network - Google Patents

Description

  The present invention relates to a collision avoidance method by wavelength allocation control in a synchronous optical packet switching network, a collision avoidance system thereof, and the like.

In an optical packet switch network (OPS), packet contention (packet collision) is an important problem, which greatly affects system performance. Collisions inevitably occur when two or more optical packets having the same wavelength are going to leave the same output port at the same time at the switch node of the network. In order to resolve the collision, several methods such as wavelength conversion, optical buffering using optical delay fiber, and detour routing have been proposed (for example, “S. Dixit, IP over WDM: building the next-generation”). optical Internet (John Wiley & Sons, 2003) "(Non-Patent Document 1). However, wavelength converters and lasers for wavelength conversion are expensive and complicated. Expected to continue in the future.
S. Dixit, IP over WDM: building the next-generation optical Internet (John Wiley & Sons, 2003).

  It is an object of the present invention to provide a wavelength allocation optimization calculation method and a collision avoidance method by wavelength allocation control in a synchronous optical packet switching network based on a new concept, which is low in cost but practical.

  The present invention basically includes (1) an organizing process for organizing a wavelength competition relationship, and (2) an optimization process for performing wavelength allocation priority optimization for minimizing wavelength competition between transit packets. (3) Collisions in the (synchronous) optical packet switching network are avoided by a process including a dynamic control process that performs dynamic wavelength allocation control based on the system load. That is, it includes refinement, optimization, and dynamic control steps (ROD). Therefore, in this specification, the method and system proposed by the present invention are also referred to as ROD.

  More specifically, the invention according to the first aspect of the present application is an optimization process for organizing the wavelength competition relationship and optimization for performing wavelength allocation priority optimization that minimizes wavelength competition between transit packets. The present invention relates to a collision avoidance method used in an optical packet switching network including a process and a dynamic control process for performing dynamic wavelength allocation control based on a system load.

  In a preferred aspect of the invention according to the first aspect of the present application, the organizing step is a competitive relationship examination step of determining whether packets having different transmission nodes are in a competitive relationship using the same wavelength in the same link of optical packet network switching. It is a method including.

  In a preferred embodiment of the invention according to the first aspect of the present application, the optimization step includes a traffic distribution information transmission step for transmitting traffic distribution information to the optical packet switch, and the optical packet switch is based on the traffic distribution information. And an optimum wavelength determination step for rearranging the wavelengths in an optimum order and prioritizing the wavelengths.

In a preferred aspect of the invention according to the first aspect of the present application, the dynamic control step is selected at random for packets that are determined to have no competitive relationship in the organizing step. For packets that have been determined to have a competitive relationship in the organizing step, the wavelength sorted in the optimization step is selected in order of priority, and the available wavelengths are examined in that order. And assigning wavelengths.

  The invention according to the second aspect of the present application is directed to organizing means for organizing wavelength competition, optimization means for performing wavelength allocation priority optimization that minimizes wavelength competition between transit packets, and system load. And a dynamic control means for performing dynamic wavelength allocation control based on the above, a collision avoidance system used in an optical packet switching network.

  A preferred embodiment of the invention according to the second aspect of the present application is a router including the above system.

  A preferred embodiment of the invention according to the second aspect of the present application is an optical network including the above system.

  As shown in the simulations described later, ROD is based on conventional pure-priority wavelength assignment (p. WANG, H. Morikawa and T. Aoyama, "Priority-based wavelength assignment for burst switched WDM optical networks," IEICE. Compared with Trans. Commun., E86-B, 5, 1508-1514 (2003).), The performance of the system is significantly increased. This system has a performance equivalent to that of a network based on a conventional wavelength conversion method, particularly when the packet loss rate is required to be 0.01 or less. Therefore, according to the present invention, it is possible to provide a high-performance system with reduced cost and a method used in the system.

  FIG. 1 is a schematic diagram illustrating the node architecture of the present invention. The optical part of the switch and the header extraction part of the input port separate and extract a small amount of the input light in order to decode the header. The header processing circuit decodes the preamble at the beginning of the packet and reads the header information. The header processing circuit informs the control device of timing information of packets input for switching and synchronization. The header modifier of the output port organizes the packet header.

  In a slot (synchronous) network, all packets have the same size. Fixed size time slots have a payload and a header. The time slot is longer than the entire packet in order to provide a guard time. All input packets arriving at the input port must be adjusted to the same phase before entering the switch. The input packet synchronizer and the output packet synchronizer are used to line up all packets input from the input port and output from the output port. In the system initialization phase, synchronization is performed to compete for delay differences between different inputs, and the situation is maintained throughout system operation time.

  Although not necessarily required in the present invention, an optical fiber delay line and a wavelength converter may be used as additional elements. For example, the optical fiber delay line buffer and the wavelength converter may be arranged to share within the node. Then they can be accessed from each other's input ports.

OPS networks are expected to perform packet-based switching optically. However, other types of interfaces need to be used to form end-to-end connections. Other networks (especially client networks) are often electrical networks. This means that an interface is required at the “edge” of the OPS network. When a packet arrives from a client network (electrical network), it needs to be converted to an optical format before being sent to the OPS network. This transformation, Ru run on the client interface. In the electro-optic interface portion of the switch, several initial first-in first-out (FIFO) buffers are included. Each buffer stores packets destined for the same destination. A burst mode transmitter and receiver are provided in the client network interface. They perform electrical-light and optical-electrical conversion at high speed. They are preferably adjustable at high speed. The time used to adjust them is preferably short compared to the duration of the slot. Burst mode transmitters and receivers that can be adjusted at very high speeds are very expensive, so the number of transmitters and receivers at each node may be reduced.

  The controller is a multifunctional control device that operates electrically. The controller preferably observes the header portion of each incoming packet and monitors the state of all wavelengths at the output port of the switch. Based on this information, wavelength allocation and packet transfer schedules are calculated, packet collisions are resolved, switches are controlled, and packet headers are replaced.

The node architecture of FIG. 1 can be used as an IP router, for example. The system of the present invention includes an organizing means for organizing the wavelength competition relationship, an optimization means for performing wavelength allocation priority optimization for minimizing wavelength competition between transit packets, and a dynamic based on the system load. A collision avoidance system used in an optical packet switching network including dynamic control means for performing appropriate wavelength allocation control. The system organizes hand stage organize wavelength competitive relationship, optimizing means performs wavelength assignment priority optimization to minimize the wavelength conflicts between transit packets, Dynamic dynamic control means based on the load of the system Wavelength allocation control. This avoids collisions in the optical packet switching network as described above.

[Optical packet routing equipment]
An optical packet routing device comprising: an O / E converter for converting an optical packet input to the optical packet routing device into an electrical signal; and the optical signal based on the electrical signal converted by the O / E converter. Based on information related to a header part of an optical packet detected by the header part detecting means for detecting information included in the header part by detecting a header part of the optical packet input to the packet routing device An optical label generating means for generating an optical label corresponding to the header portion of the optical packet, and a serial / parallel converting means for converting the electric signal converted by the O / E converting means into a parallel signal; A transmission speed adjusting means for adjusting a transmission speed of the electric signal converted by the serial / parallel converting means; and the transmission speed adjusting means. An E / O conversion means for converting an electrical signal whose transmission speed is adjusted by an adjustment means into an optical signal, an optical label generated by the optical label generation means, and an optical signal converted by the E / O conversion means This is realized by an optical packet routing device comprising an optical signal generating means for generating a combined optical signal.

[Optical packet switching network]
FIG. 2 is a diagram showing a basic configuration of an optical packet switching network. As shown in FIG. 2, the optical packet switching network (21) includes a backbone network (23) including a backbone node (22), and a metro network (24) which is a network for optical packet communication outside the backbone network. , And an edge node (25) connecting the backbone network and the metro network.

  Then, the edge node (25) includes an ingress edge node that is an optical packet routing device when an optical packet enters the backbone network (23) from the metro network (24), and an optical packet that transmits the optical packet to the backbone network (23 ) To the metro network (24), and a leaving edge node which is an optical packet routing device.

  In the metro network (24), for example, optical information is transmitted using optical packets having a single wavelength. The optical packet includes a header portion including information such as a destination and a payload portion including data regarding contents to be transmitted.

  The network of the present invention employs the above-described system as an optical packet routing device, for example, so that collision can be avoided with favorable performance at low cost.

[System Model] In order to verify the method and system according to the present invention, the following system was used as a model. Note that all timing control and the like may be performed by a controller such as a controller. The control unit may be implemented by hardware such as a programmable logic device. Further, it may be configured by software loaded with a predetermined control program, a CPU, a calculation unit, a memory, an input / output unit, a bus connecting each element, and the like. The CPU reads a control program in response to an input signal from the input unit, receives a command from the control program, reads out a signal input from the input unit and various information stored in the memory, and outputs various information to the arithmetic unit. It is only necessary to perform the calculation, store the calculation result in the memory, and output a control command or the like as appropriate from the output unit. Examples of the control program include a program for realizing each step disclosed in the present specification, a program for extracting or adjusting header information, a program for controlling the timing of various signals, and the like. N optical packet switches are connected to the bidirectional fiber link to construct an optical packet switch network (OPS) of arbitrary topology. Each fiber contains M wavelengths. An IP client network is connected to the edge of the OPS network. At the edge node, many buffers (subqueues) are input. Each buffer (subqueue) is a packet for the same destination. The packet length of each buffer is represented as Q, which is in packet units. Each node has A wavelength-tunable transfer devices and a burst receiver at 10 Gb / s. The tuning time is assumed to be shorter than the slot duration. In slotted networks, packets are assumed to be aligned before being input to the switching node, and are fixed at 12,000 bits. For simplicity, it is assumed that each packet is placed in one 1.2 microsecond slot (for detailed node configuration, see Non-Patent Document 1 “S. Dixit, IP over WDM: building the next-generation optical Internet (John Wiley & Sons, 2003). ”).

Assume that each node can observe all wavelengths in the slot and the state of all transferred objects (free or busy) on a slot basis. This is called lambda monitoring (A. Carena, et al., "RingO: an experimental WDM optical packet network for metro applications," JSAC, 22, 8, 1561-1571 (2004).). Collision between forward packets and the local input packet, the local input packet buffer (subqueue), shall be avoided by electrically input. Packet transfers between buffers (subqueues) shall be planned in a “round robin” manner. A simple round robin drop method is used when a transfer packet contention or a receiver contention occurs. To simplify the discussion and concentrate the focus on wavelength allocation, the packet route is fixed to the shortest path.

[Organization of wavelength competition]
First, we organize the wavelength competition in order to use limited wavelength resources effectively. In order to sort out the wavelength competition, basically, by verifying the route, it is determined that there is no possibility of collision and it is determined that it is not necessary to assign a different wavelength to the packet. For packets that may collide, different wavelengths are assigned as much as possible. This wavelength assignment is performed after calculating the possibility of collision. These controls may be performed according to a control signal from a control unit such as a controller. In this model, the graph representing the collision is G _C = (V _C , E _C ). Here, each node in the graph V _C corresponds to a virtual path in the network. E _C , which indicates a bidirectional link between two adjacent nodes, represents a wavelength competition that requires different wavelengths to avoid potential wavelength contention problems. This competitive relationship is represented by a matrix [g _ij ]. If there is a conflict between virtual paths i and j, [g _ij ] is 1; otherwise [g _ij ] is 0. The process of organizing the competitive relationship is shown in FIG. 3 and FIG. FIG. 3 is a pseudo code showing wavelength competition arrangement. FIG. 4 is a conceptual diagram for explaining the wavelength competition relationship after arrangement. 4 (a) shows a network of pre-organize, FIG. 4 (b) shows the post-organizing conflict graph G _C (conflict graph).

The purpose of organizing the wavelength competition is to minimize the wavelength competition among all packets. The collision relationship means a wavelength competition relationship that can occur between packets having different combinations of source and destination nodes. Packets can collide when the same wavelength is present in the same fiber at the same time. In general, the collision relationship between different packets can be represented by a graph. Let this graph be G _p = (V _p , E _p ). V _p represents a set of vertices, and | V _p | = N. E _p indicates a set of links. This situation is shown in FIG. 4 (a) representing the network. The vertex V _p (0-3) in the graph of FIG. 4A corresponds to a node in the network. The end between the two vertices corresponds to the fiber between the two nodes. Since the packet route corresponds to the virtual path in the graph G _p , the combination of routes in the network corresponds to the combination of virtual buses in the graph G _p . Then, an arbitrary collision graph G _C = (V _C , E _C ) as shown in FIG. 4B can be designed. If, when two virtual paths share certain edges in the graph G _p, collision relationship occurs between the two virtual paths. Corresponding nodes in the graph G _C are connected by an edge different from the traveling direction. If the graph of FIG. 4A is not rearranged and converted as it is, a graph showing the collision relationship between virtual paths as shown in FIG. 4C is obtained.

That is, the wavelength competition relationship may be arranged as follows. The shortest path is calculated from the combination of the source note and the target node of the optical packet (step 1). This calculation can be accomplished by known methods. A collision graph G _C is created from the graph G _p indicating the original network (step 2). To create a conflict graph G _C from the graph G _p may be grasped and the vertex corresponding to each node included in the graph G _p, a route connecting the vertices can be created easily conflict graph G _C. The conflict graph G _C is expressed by a matrix [g _ij ] for this competitive relationship for all node combinations (i, j). Note that all [g _ij ] are initialized so that the initial value is zero. Each node of the combination of conflict graph G _C corresponding to the combination of the virtual path (i, j) and X is a set of links for the virtual path corresponding to index i for the link for the virtual path corresponding to the index j Y, which is a set of (1), is prepared (step 3). Elements included in X and Y are links along virtual paths, respectively. Thereafter, the wavelength relationship is optimized using a pseudo code as shown in FIG.

That is, if X and Y do not contain the same element, [g _ij ] is 0, ie no collision occurs; and the first link of X is included in Y, or the first link of Y Is also included in X, [g _ij ] is set to 0, that is, no collision occurs; otherwise, [g _ij ] is set to 1, that is, a collision can occur.

  More specifically, in order to sort out the wavelength competition, each node included in a certain network is grasped. Preferably, each node is labeled. Then, the route between each node is grasped. Next, the source (departure) node and the target node of a plurality of optical packets are grasped. Then, the shortest path of the optical packet is grasped. At that time, it also knows which node it goes through. Then, it is determined whether or not a collision with another optical packet can occur. That is, since a collision can occur when optical packets having the same wavelength enter the same route (fiber) at the same time, it is first determined whether the shortest paths include a common path. If there is no common path, no collision occurs. Also, for example, if two optical packets are considered, it is assumed that the source nodes of the two optical packets are different. Therefore, if the first path of one of the optical packets is included in the partner path, the collision is Suppose that it does not occur. That is, in the step of organizing the wavelength competition relationship according to the present invention, the above-described examination is performed for each optical packet, and it is desirable that wavelength collision cannot occur in the shortest path for all the optical packets.

[Wavelength Assignment Priority Optimization]
Next, based on the organized wavelength competition, limited wavelength resources are allocated to minimize potential wavelength collisions. Through this optimization process, a wavelength priority list is created corresponding to each virtual path. In each virtual path, M wavelengths are prepared for the next priority. The wavelength with the highest priority is most preferred because it is least likely to cause a collision. In this wavelength allocation optimization step, traffic dispersion information is preferably used. By using Extended Interior Gateway Protocols such as GMPLS OSPF-TE / IS-IS-TE, traffic distribution information can be reported to each optical packet switch. For simplicity, it is assumed that traffic information is known here.

The optimization and subroutine processes are shown in FIGS. 5 and 6, respectively. FIG. 5 is a pseudo code showing an optimization process, and FIG. 6 is a pseudo code showing a subroutine process. Reference numerals p and p ′ denote virtual paths. w represents a wavelength. t _p indicates the traffic amount of the virtual path p. The set of P includes all virtual paths that have a competitive relationship. P _l is a subset of P and includes a virtual bus through link l. Lp indicates a list of links along the virtual path p. Ψ represents a set of all wavelengths. Ψ denotes a set of wavelengths including M wavelengths having indices from 1 to M. Ψp is a set including the assigned wavelength of the virtual path p. Ψp is included in Ψ. Δ ^w _l is updated after each optimization of the subroutine, and is used as a steady state until the next optimization. traffic load variable). λ ^w ( _{p, l} ) represents a binary number indicating wavelength assignment in the virtual path p passing through the link l, and p is included in P _l . Then, when the virtual path p passes the wavelength w through the link l, λ ^w ( _{p, l} ) is 1, and 0 otherwise. θ ^w _p indicates a binary number that is 1 when the wavelength w is assigned to the virtual path p, and is 0 in other cases. Equation (5) in FIG. 6 indicates the possibility of collision of the wavelength w in the virtual path p. In order to avoid collision, a wavelength w having a small C ^w _p is selected.

In order to optimize the wavelength allocation priority, specifically, the following may be performed.
First, initialization is performed (step 1). That is, all of the virtual path that does not contain anything Ψp for, for all wavelengths w and all links l, variable traffic volume delta ^w _l to 0. Let λ ^w _{(p, l)} be zero. θ ^w _{p is set} to 0. This step is expressed as follows.

  For all wavelengths w on all links l, a wavelength w with a small cumulative traffic volume is selected. The amount of traffic for all wavelengths w of all links l is summed. That is, the priority of wavelength allocation is calculated according to the following subroutine. Record the selected wavelength for each virtual path in a single calculation. The constraint is set so that the wavelength once selected is not selected again in the next calculation. For the next wavelength allocation, the traffic volume at each wavelength is updated (step 2). This step is expressed as follows.

  Next, the release and constraints are stored (step 3). This step is expressed as follows.

  Return to step 2 and repeat it M times (step 4). The optimization is finished as described above (step 5).

The meaning of each step will be briefly described. Step 1 is an initialization process. Step 2 is preferably a process for calculating a wavelength with the following subroutine and selecting a wavelength with a low possibility of collision. The target evaluation value is the above formula (5). Step 3 is a process for recording one calculation result in Step 2 and assigning one wavelength w to each virtual path. Thereafter, Δ ^w _l is updated, and λ ^w ( _{p, l} ) is reset for the next calculation. Another wavelength may be selected by performing step 2 again on the updated Δ ^w _l and calculating the collision probability. The wavelength selected at this time is not very optimal. However, it is preferable to store the selected wavelength. In this way, a list of wavelength assignments can be obtained for each virtual path by repeating Steps 2 and 3 M times. Note that the wavelength obtained by the last calculation is the most congested wavelength.

  Step 2 above may be the following subroutine for optimization. That is, the following inductive subroutine.

The virtual paths included in P are sorted in descending order according to the number of hops (step 1). However, for all w 1 , p and l, the initial value is 0 for λ ^w _{(p, l)} .

The virtual path p is selected from the beginning, and the wavelength w is selected for the virtual path having the smallest collision coefficient C ^w _p in equation (1) (step 2). However, all w are included in Ψ−Ψp. This wavelength satisfies the minimum value of Equation (5) ^{C w} _p.

When w is a selected wavelength of the virtual path p, λ ^w _{(p, l) is set} to 1 and θ ^w _{p is set} to 1 for all l included in L _p (step 3).

  Returning to step 2, the optimization is repeated until wavelengths are assigned to all virtual paths included in P (step 4). After this is finished, the subroutine is finished (step 5).

[Dynamic Wavelength Assignment Control Based on Traffic Volume] In a pure PWA, if a desired wavelength cannot be obtained, a candidate with the next highest priority is assigned immediately if it can be obtained. In fact, the highest priority wavelength is not always obtained, and the most preferred wavelength is not always assigned to a packet. Therefore, the conventional wavelength allocation optimization method is not always efficient. The present invention improves the preferred wavelength utilization for each packet by holding the packet and waiting until the preferred wavelength is obtained. Since the packet has a competitive relationship with other packets, the wavelength is selected from the wavelength with the highest priority in the wavelength selection space. If the number of packets contained in the buffer increases, for example, under higher loads, the wavelength selection space is expanded or otherwise reduced. The size of this wavelength selection space can be determined by the function T (x). Here, x (0 <x <Q) indicates the current number of packets in the buffer. T (x) is the size of the wavelength selection space, as shown in Equation (2). In equation (2), operation └X┘ is a function that is the largest integer not greater than X. For packets that do not have a competitive relationship with other packets, the wavelength is randomly selected from the M wavelengths.

  Packets are arranged in order of priority. When the number of packets in the buffer is smaller than Q / M, the wavelength search space is 1. That is, only the first packet in the buffer is assigned as the highest priority. On the other hand, if the number of packets in the buffer increases, for example, if Q / M <x <2Q / M, the wavelength search space is 2, and if it is available, the first packet in the buffer is the most. Assigned as high priority, otherwise assigned if the second highest priority wavelength is available. As the number of packets in the buffer increases, that is, as the packet load increases, the wavelength search space increases. If (M-1) Q / M <x, all M wavelengths are searched for assignment. Wavelengths are assigned and examined in order of priority. On the other hand, when the packet load decreases, the number of packets in the buffer also decreases. And since the collision relationship with other wavelengths concentrates on assigning the most preferable wavelength, the wavelength search space gradually becomes smaller. Such a mechanism can be effectively used to hold packet transfers and assign better wavelengths. And such a mechanism reduces the occurrence of collisions in the optical part of the network and the leaving edge node. Furthermore, the wavelength search space is dynamically adjusted according to the system load. That is, if the system load changes, the wavelength allocation policy (method), that is, the wavelength search space is converted accordingly. On the other hand, packets between a transmission source and a destination that are not related to collisions with other packets are randomly assigned from available wavelengths.

[simulation result]
The simulation was performed based on the same NSFNET topology as disclosed in Non-Patent Document 1. The fiber length of each link was assumed to be 36 km (approximately 100 time slots). The traffic was uniform and the packet arrival was assumed to be a Poisson process. In this simulation, ROD was compared with three schemes. That is, (1) random wavelength allocation (Rand); (2) pure PWA (pure PWA) and (3) complete wavelength conversion method (Conv). These three schemes do not perform dynamic wavelength allocation control. Compare packet loss rate to average packet transmission rate (ie, the number of bits of packets transmitted per unit time divided by link speed at each transmitter) from the viewpoint of average delay did. At that time, the parameters M, Q and A were changed. Each case with different parameters was labeled “Scheme-MaQbAc”. Here, a indicates the number of wavelengths, b indicates the length of the buffer, and c indicates the number of transmitters or receivers. Some of the results obtained are shown below.

  FIG. 7 shows the packet loss rate when M = 16, Q = 100, and A = 8. On the other hand, FIG. 8 shows the average delay when M = 16, Q = 100, and A = 8. Table 1 shows the upper limit of the average packet transmission rate when the allowable packet loss rate is 0.01. When the load is light, ROD has better performance than the case of wavelength conversion (Conv) because the dynamic wavelength allocation control packet can wait in the buffer for a preferred wavelength that can reduce collisions in the network. Therefore, ROD can achieve a low packet loss rate. As shown in Figure 8, the delay performance is sacrificed, but the delay caused by queuing is negligible compared to the long-distance link propagation delay, which is the largest factor when considering the delay. . Furthermore, it was confirmed that the preferred buffer length for ROD is not so long. That is, when the buffer length is considerably increased, the system cost increases, but the ROD performance is not improved so much.

  The system of the present invention uses a local buffer in the electrical portion of the network based on the above arrangement and wavelength allocation priority optimization process to avoid wavelength access and collisions. Then, collisions in the optical part of the network are alleviated, and a dynamic wavelength allocation control scheme can be applied to balance the connection problem between the electrical part and the optical part of the network. Simulations have shown that ROD significantly improves system performance. Even when a packet loss rate of 0.01 or less is required, ROD can be said to be a basic system that can reduce the cost in wavelength conversion.

  The present invention can be applied not only to an asynchronous optical packet switching system but also to an optical path network wavelength allocation technique using a control technique such as GMPLS. That is, the present invention can be used in fields such as optical information communication.

FIG. 1 is a schematic diagram illustrating the node architecture of the present invention. FIG. 2 is a diagram showing a basic configuration of an optical packet switching network. FIG. 3 is a pseudo code showing wavelength competition arrangement. FIG. 4 is a conceptual diagram after arrangement. FIG. 4A shows the network after arrangement, and FIG. 4B shows the collision graph after arrangement. FIG. 4 (c) shows a collision graph that is not organized. FIG. 5 is a pseudo code showing the optimization process. FIG. 6 is a pseudo code showing the subroutine process. FIG. 7 shows the packet loss rate when M = 16, Q = 100, and A = 8. FIG. 8 shows the average delay when M = 16, Q = 100, and A = 8.

Explanation of symbols

21 Optical packet switching network
22 Core node
23 Core network
24 Metro network
25 Edge node

Claims (6)

Organizing means for organizing wavelength competition, optimization means for obtaining wavelength allocation priority optimization information that minimizes wavelength competition between transit packets, and dynamic wavelength allocation control based on system load A transit packet collision avoidance method in an optical packet switching network using a transit packet collision avoidance system comprising dynamic control means for performing,
An organizing process for organizing the wavelength competition,
An optimization process for obtaining wavelength allocation priority optimization information that minimizes wavelength competition between transit packets;
A dynamic control process for performing dynamic wavelength allocation control based on system load;
Including
The organizing process includes:
The organizing means includes a competitive relationship examination step of determining whether packets having different transmission nodes are in a competitive relationship using the same wavelength in the same link of optical packet network switching, and the optimization step includes
A traffic distribution information transmission step in which the optimization means receives traffic distribution information relating to traffic volumes of a plurality of paths included in the optical packet network;
The optimization unit obtains a collision probability value for each wavelength based on the traffic distribution information, and re-arranges the wavelengths in ascending order of the calculated collision possibility value to prioritize the wavelengths. And
The dynamic control process includes:
The dynamic control means assigns a wavelength randomly selected from the available wavelengths to a packet that is determined not to have a competitive relationship in the organizing step,
For packets determined to have a competitive relationship in the organizing step, the dynamic control means allocates the number of wavelengths aligned in the optimization step according to the system load ,
The dynamic control process includes:
The dynamic control means assigns a wavelength randomly selected from the available wavelengths to a packet that is determined not to have a competitive relationship in the organizing step,
For packets that are determined to have a competitive relationship in the organizing process,
The dynamic control means comprises:
Obtaining information about the number x of packets contained in a buffer in the system corresponding to the load of the system;
When Q is a packet length that is the maximum number of packets that the buffer can contain, and M is the number of wavelengths that can be allocated in the system,
Find the magnitude relationship between x and Q / M and 2Q / M,
if x is less than Q / M, assign the first packet in the buffer to the highest priority wavelength arranged in the optimization step;
When x is greater than or equal to Q / M and less than 2Q / M, if the highest priority wavelength arranged in the optimization process is available, the first packet in the buffer is arranged in the optimization process. If the highest-priority wavelength assigned in the optimization step is not available, the second packet arranged in the optimization step is the second packet arranged in the buffer. A method for avoiding transit packet collisions in an optical packet switching network , including a step of assigning wavelengths having high priority .
The competitive relationship examination step receives information on a starting node and a destination node of the first packet, obtains a shortest path and a transit node from the starting node to the destination node of the first packet, and determines the first packet. Information on the starting node and the destination node of the second packet that are different from each other and the sending node is obtained, the shortest path and the transit node of the second packet are obtained, and the shortest path and the second shortest path of the first packet are obtained. Determining whether or not a common path is included, and determining that the first packet and the second packet are not in a competitive relationship if there is no common path; and an initial path of the first packet Is included in the shortest path of the second packet, and if included, it is determined that the first packet and the second packet are not in a competition relationship; It is determined whether or not the first path of the second packet is included in the shortest path of the first packet. If included, it is determined that the first packet and the second packet are not in a competitive relationship. The method for avoiding transit packet collisions in an optical packet switched network according to claim 1, further comprising:
Organizing means for organizing wavelength competition,
An optimization means for performing wavelength allocation priority optimization that minimizes wavelength competition between transit packets;
Dynamic control means for performing wavelength allocation control;
A transit packet collision avoidance system for use in an optical packet switching network comprising:
The organizing means receives information on the starting node and the destination node of two or more packets having different transmitting nodes, obtains the shortest path of each packet, and packets having different transmitting nodes have the same wavelength on the same link of optical packet network switching Determine if there is a competition to use,
The optimization means receives traffic distribution information related to the traffic amount of a plurality of paths included in the optical packet network, determines a collision probability value for each wavelength based on the traffic distribution information, and determines the calculated collision probability. Reorder the wavelengths in ascending order of values to prioritize the wavelengths,
The dynamic control means comprises:
For the packet that the organizing means determines that there is no competition, a wavelength randomly selected from the available wavelengths is assigned,
For packets determined to be in a competitive relationship by the organizing means, a number corresponding to the load on the system is assigned according to the priority order from the wavelengths arranged by the optimizing means ,
The dynamic control means includes
For the packet that the organizing means determines that there is no competition, a wavelength randomly selected from the available wavelengths is assigned,
For packets that the organizing means determines to have a competitive relationship,
Obtaining information about the number x of packets contained in a buffer in the system corresponding to the load of the system;
When Q is a packet length that is the maximum number of packets that the buffer can contain, and M is the number of wavelengths that can be allocated in the system,
Find the magnitude relationship between x and Q / M and 2Q / M,
If x is less than Q / M, assign the highest priority wavelength arranged by the optimization means to the first packet in the buffer;
When x is greater than or equal to Q / M and less than 2Q / M, when the wavelength with the highest priority arranged by the optimization means is available, the optimization means arranges the first packet in the buffer. If the wavelength with the highest priority is assigned and the wavelength with the highest priority arranged by the optimization means is not available, the second packet arranged by the optimization means is assigned to the first packet in the buffer. Assign a higher priority wavelength,
A collision avoidance system for transit packets used in optical packet switching networks.
The organizing means receives information relating to a starting node and a destination node of the first packet, obtains a shortest path and a transit node from the starting node to the destination node of the first packet, and transmits the first packet and the transmission node. The node receives information about the starting node and the destination node of the second packet, and finds the shortest path and the transit node of the second packet, and is common to the shortest path and the second shortest path of the first packet. It is determined whether or not a path is included. If there is no common path, it is determined that there is no conflict because the first packet and the second packet do not collide, and the first path of the first packet is It is determined whether the second packet is included in the shortest path, and if it is included, it is determined that the first packet and the second packet are not in a competitive relationship, and the second packet Determining whether the first path of the packet is included in the shortest path of the first packet and, if included, determining that the first packet and the second packet are not in a competitive relationship, 4. The transit packet collision avoidance system according to claim 3 , wherein packets having different transmission nodes determine whether or not there is a competitive relationship using the same wavelength in the same link of optical packet network switching.
A router comprising the system according to claim 3 or 4 .
An optical network including the system according to claim 3 .

Images