JP2014160920A - Tdm network system and scheduling method for the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a TDM network system capable of performing dynamic resource allocation at high speed and increasing its scale.SOLUTION: A scheduler 20 comprises: a traffic variation difference calculation unit 31 for obtaining variation difference information by comparing a current required time slot matrix with a previous required time slot matrix; a decrease amount removal calculation unit 33 for deleting an already allocated time slot with regard to a path whose traffic decreases from a link schedule table 25, on the basis of the variation difference information; and an increase amount pasting calculation unit 35 for additionally allocating a time slot at an empty position in the link schedule table 25 with regard to a path whose traffic increases, on the basis of the variation difference information.

Description

  The present invention relates to a network system based on TDM (time division multiplexing) technology, and more particularly to a system that dynamically allocates resources and a method for dynamically allocating resources, that is, a scheduling method.

  In a TDM network, after a plurality of nodes share a transmission path, each node determines a timing for transmitting data to the transmission path, and each node transmits data at a periodic timing assigned to itself. Send out on the transmission line. Normally, a period of a certain time length is defined as one time slot (TS: timeslot) as a transmission timing, and a time slot used for transmission is determined for each combination of a transmission source node and a destination node. A combination of a source node and a destination node is generally called ground. Allocating a time slot for data transmission for each ground is called scheduling. At this time, time slots are assigned for each ground so that they do not overlap with other grounds, and more time slots are assigned to grounds with a large required bandwidth.

  FIG. 1 shows a network composed of a ring-shaped transmission line as an example of a TDM network. In this network, five nodes 11 appearing in the drawings A to E are provided on a ring-shaped transmission line. Here, it is assumed that the transmission path is an optical transmission path that transmits a signal in one direction. In the transmission line, a section sandwiched between two adjacent nodes 11 is called a link 10. For example, the link 10 having both ends of the node A and the node B is represented by a in the figure, and there are links b to e in order. In the illustrated network, the node A is provided with a scheduler 20 that schedules time slots in the entire network.

  In order to perform data transmission efficiently, that is, without wasting the bandwidth of the transmission path, it is necessary to allocate time slots without gaps. A table representing the status of air slot blockage defined for each link in the network is called a link schedule table. Therefore, it can be said that scheduling is an operation of allocating communication to each time slot indicated on the link scheduling table.

  FIG. 2 shows one shown in Non-Patent Document 1 as an example of a conventional scheduling method in a TDM network.

  First, in step 801, the traffic for each ground is totaled, and then in step 802, it is converted into the number of necessary time slots (number of required time slots). At this time, the number of time slots is obtained by converting the traffic between the grounds into an integer ratio. However, there is a constraint that the traffic allocated to each link must not exceed the link capacity. From this restriction, immediately for each node, the restriction that the sum of the transmission traffic should not exceed the link capacity when that node functions as the source node, and the reception that occurs when that node functions as the destination node. There is also a constraint that the total traffic must not exceed the link capacity. A matrix having one of the transmission source node and the destination node as a row and the other as a column and the number of time slots for each ground as an element is referred to as a TS (time slot) matrix.

  Next, in step 803, a TDM control repetition period is calculated, and a link schedule table is created by assigning each ground to the number of required time slots for each empty time slot. For example, in the technique described in Non-Patent Document 1, a superframe length that can accommodate all time slots is calculated. In this process, the time slots that will be accommodated in the via-to-ground link are accumulated for each link, and among the accumulated values for each link, the maximum accumulated value is obtained through all the links, and the superframe is based on this. Determine the length. Then, a link schedule table is created by assigning each ground to the number of required time slots for empty time slots corresponding to the superframe length. Here, the repetition period of the TDM control is calculated. However, it is also possible to determine the repetition period in advance and perform allocation to an empty time slot based on the determined period.

  There are various allocation policies when allocating to free time slots. For example, there are first fit allocation in which allocation is performed immediately when a free space is found, and continuous TS priority allocation in which allocation is performed immediately if the required number of time slots can be secured continuously.

  Next, in step 804, a node schedule table used for actual processing at each node is created from the link schedule table. The node schedule table indicates information for instructing the timing of route switching and data transmission in each node, and can be generated by conversion from the link schedule table.

  Finally, as shown in step 805, the processing of the scheduling is completed by transmitting the processing schedule of the node to each node based on the link schedule table.

  Here, assuming that the number of nodes is N and assuming full mesh communication, the number of paths P, that is, the total number of grounds that can be assumed is P = N (N−1). For example, if the number of nodes is 5, P = 20. If it is a unidirectional ring transmission line, the number of links is also N, and P / 2 paths pass through each link.

  The scheduling process described above is performed by the scheduler 20 provided in any one of the nodes 11. FIG. 3 is a block diagram showing an example of a configuration of a conventional scheduler.

  The scheduler 20 collects AC traffic volume from each node 11 and distributes a link schedule table to each node 11 to realize dynamic control of the entire network. The scheduler 20 includes an AC traffic volume totaling unit 21, a path selector 23, a table writing unit 24, a table conversion unit 26, a table transmission unit 28, and a timer 29. Further, the table 20 includes an unallocated path table 22, a link schedule. A table 25 and a node schedule table 27 are held.

  The AC traffic counting unit 21 obtains traffic information periodically reported from each node 11 in the network (for example, a buffer accumulation amount for each destination of each node), and from this traffic information, The requested time slot number is calculated and written in the unallocated path table 22. Since the unassigned path table 22 holds a path to which time slots should be assigned and the number of time slots to be assigned, the path selector 23 sets the required time slot number stored in the unassigned path table 22. Based on this, a path is selected based on a predetermined allocation policy, and the table writing unit 24 writes the path selected by the path selector 24 to the link schedule table 25 as a time slot allocation, thereby assigning a time slot to the path. I do. The table conversion unit 26 converts the link schedule table 25 into a node schedule table 27 for each node 11, and the table transmission unit 28 transmits the processing schedule of the node to each node 11 based on the node schedule table 27. The timer 29 is for performing time management by synchronizing with each node 11. In the scheduler 20, the path selector 23 and the table writing unit 24 sequentially execute a process of selecting a path for time slot allocation and performing table writing.

  As an example of a TDM network to which such time slot scheduling is applied, there is one in which a host computer 12 is provided for each node 11 as shown in FIG. In this TDM network, consider a case where the host computer 12 transmits data to a host computer (another host) connected to another node. In this case, as shown in FIG. 5, first, in step 851, when the host computer 12 transmits data addressed to another host, the host computer 12 is connected by the inter-host data signal (step 852). To the existing node 11. In step 853, the node 11 receives the data, and performs buffering of the data and monitoring of the traffic. As a result, traffic information between the hosts is sent from the node 11 to the scheduler 20 (step 854). In step 855, the scheduler 20 performs scheduling as described above and notifies the scheduling result. The scheduling result is sent to each node 11 as a processing schedule (step 856), and each node 11 performs data transfer according to the schedule in step 857. As a result, the data addressed to the other host buffered in the node 11 is sent to the destination host computer 12 as an inter-host data signal (step 858).

  As a scheduling method in a TDM network including a ring-shaped transmission path, there is a method of Zhang et al. (Non-patent Document 2) in addition to the method described in Non-Patent Document 1 by Gohaku et al. In these, all given traffic can be accommodated, sorting and condition determination are performed according to a certain criterion, and allocation is performed sequentially for each pass. Here, it means to search and secure an empty slot position that is continuous on the via link (in the link schedule table, time slots at the same time are continuous across links). Wen et al. Are also examining resource allocation for an optical TDM network in a general network (Non-patent Document 3).

  FIG. 6 shows an example of scheduling by the method described in Non-Patent Document 1 when the number of nodes is 5 (nodes A to E) and the network is a one-way ring network. This method is a method called First Fit among the methods based on the allocation policy described above. S1 to S9 represent time slot numbers. In this method, as shown in the required TS matrix on the left side of the figure, it is assumed that the number of required time slots for each path (for each ground) is given, and each path is empty in descending order of path length and in ascending order of time slot numbers. Search for and assign a time slot. Since the number of nodes is 5, the longest path length is 4. Therefore, slots are allocated in order from the A → E path and the B → A path having a path length of 4.

  By the way, in the conventional scheduling method as described above, when dynamic bandwidth reallocation or link scheduling table rewriting is performed to follow traffic fluctuation, as described above, if the number of nodes is N, the number of grounds is Since the number of nodes increases in the order of the square of N, the amount of calculation increases accordingly. In this way, as the network scale increases, the amount of calculation required increases rapidly, and as a result, it becomes impossible to cope with an increase in the scale of the network.

  For example, as shown in FIG. 7, when there is traffic destined for nodes A and B in the case of a logical network, if the network is small, the allocated bandwidth may be changed in response to the change in the requested traffic amount. Although it is possible, if the network becomes large and the number of paths and links to be calculated increases, the calculation time cannot catch up with the traffic fluctuation period, and it becomes impossible to follow the traffic fluctuation.

K. Gokyu, K. Baba, and M. Murata, "Path accommodation methods for unidirectional rings with optical compression TDM," IEICE Transactions on Communications, vol. E83-B, pp. 2294-2303, Oct. 2000 X. Zhang and C. Qiao, "Pipelined transmission scheduling in all-optical TDM / WDM rings," in Proc. Int. Conf. Computer Communication and Networks, Sept. 1997, pp. 144-149 B. Wen, R. Shenai, and K. Sivalingam, "Routing, Wavelength and Time-Slot-Assignment Algorithms for Wavelength-Routed Optical WDM / TDM Networks," J. Lightwave Technol., Vol. 23, no. 9, pp 2598-2609, Sept. 2005

  In a TDM network, when dynamic bandwidth reallocation or link schedule table rewriting is performed in order to follow traffic fluctuations, conventionally, bandwidth reassignment or rewriting is performed for the nodes of the entire network. As the number of nodes increases and the number of nodes increases, the calculation time for reassignment becomes longer at an accelerated rate, which makes it difficult to follow traffic fluctuations.

  An object of the present invention is a TDM network system capable of performing dynamic resource allocation such as dynamic bandwidth reallocation and rewriting of a link schedule table at high speed and increasing the number of nodes to form a large-scale network. Will be offered.

  Another object of the present invention is a scheduling method in a TDM network, which can perform dynamic resource allocation such as dynamic bandwidth reallocation and link schedule table rewriting at high speed, and increases the number of nodes. The object is to provide a scheduling method that can be applied to a large-scale network.

  The TDM network system according to the present invention is a TDM network system including a plurality of nodes and links connecting the nodes. For each ground, which is a combination of a communication source node and a destination node, the TDM network system corresponds to the traffic volume on the ground. A scheduler for allocating time slots to the ground, and the scheduler includes a link schedule table indicating a time slot allocation status for each link and for each time slot position, and for each ground, the previous scheduling execution time and the current scheduling execution time Difference detection means for detecting the difference in traffic volume between the two, and stripping to delete the time slot assigned to the ground from the link schedule table in response to a decrease in the traffic volume of the ground for each ground Peeling off A paste calculation for performing a paste calculation process and adding a time slot for the ground at an empty position in the link schedule table in response to an increase in the traffic volume of the ground for each ground. And means.

  The scheduling method of the present invention includes a plurality of nodes and links connecting the nodes, and for each ground that is a combination of a communication source node and a destination node, sets a time slot in the ground according to the traffic amount in the ground. In the scheduling method in the assigned TDM network system, a step of detecting a difference in traffic volume between the previous scheduling execution time and the current scheduling execution time for each ground, and a decrease in the traffic volume of the ground for each ground In response to this, a step of performing a stripping process for deleting the time slot assigned to the ground from the link schedule table indicating the allocation status of the time slot for each link and for each time slot position, and for each ground Increased traffic volume Depending on the bets, and having a the steps of performing a paste operations to add a time slot for the ground in a position that is vacant in the link schedule table.

  In the present invention, for each ground, a difference in traffic volume between the previous scheduling execution time and the current scheduling execution time, that is, a traffic fluctuation amount is detected, and time slot allocation is performed only for the ground where the traffic fluctuation has occurred. Recalculation. As a result, the calculation time is significantly reduced as compared with the case of performing recalculation for all the grounds, and it becomes possible to perform dynamic resource allocation at high speed. Therefore, according to the present invention, dynamic resource allocation in a large-scale TDM network becomes possible.

It is a figure which shows an example of a TDM (time division multiplexing) network. It is a flowchart explaining the conventional scheduling method in a TDM network. It is a block diagram which shows the structure of the conventional scheduler in a TDM network system. It is a block diagram which shows an example of a TDM network provided with a host computer for every node. FIG. 5 is a sequence diagram for explaining scheduling processing in the TDM network shown in FIG. 4. It is a figure explaining the scheduling by the method called First Fit. It is a figure explaining the increase in the calculation time by network enlargement. It is a figure which shows the concept of the scheduling method of one Embodiment of this invention. It is a figure which shows the concept of the scheduling method of one Embodiment of this invention. It is a figure which shows an example of the partial rewriting of a link schedule table. It is a figure explaining the determination method of the order which performs a stripping process. It is a figure explaining the determination method of the slot position which performs a stripping process. It is a figure explaining the determination method of the order which performs a sticking process. It is a figure explaining the determination method of the slot position which performs a sticking process. It is a figure explaining the process of new path accommodation. It is a block diagram which shows the structural example of a node. 2 is a block diagram showing a configuration of a scheduler that executes Method 1. FIG. 11 is a sequence diagram illustrating processing of method 1. FIG. 11 is a sequence diagram illustrating processing of method 1. FIG. It is a figure explaining operation | movement of an alternating current traffic total part. It is a figure explaining operation | movement of a traffic fluctuation | variation difference calculation part. It is a figure explaining the process of a removal part removal calculation part. It is a figure explaining the process of a removal part removal calculation part. It is a figure explaining the process of a removal part removal calculation part. It is a figure explaining the process of an increase paste calculation part. It is a figure explaining the process of an increase paste calculation part. It is a figure explaining the process of an increase paste calculation part. It is a figure explaining the various examples of the determination method of the position which performs a stripping process. It is a figure explaining the various examples of the determination method of the position which performs a sticking process. FIG. 10 is a block diagram illustrating a configuration of a scheduler that executes Method 2. FIG. 10 is a sequence diagram illustrating processing of method 2. FIG. 10 is a sequence diagram illustrating processing of method 2. FIG. 12 is a block diagram illustrating a configuration of a scheduler that executes Method 3. 11 is a sequence diagram illustrating processing of method 3. FIG. 11 is a sequence diagram illustrating processing of method 3. FIG. FIG. 10 is a block diagram illustrating a configuration of a scheduler that executes a limitation on the number of stripping processes and pasting processes in Method 1; FIG. 11 is a sequence diagram for explaining processing when executing a limit on the number of stripping processing and pasting processing in Method 1; It is a block diagram which shows the structure of the scheduler which implement | achieves time reduction until control reflection. It is a block diagram which shows the structure of the scheduler which implement | achieves suppression of the deviation from a scheduling optimal solution.

  Next, a preferred embodiment of the present invention will be described with reference to the drawings.

  First, the concept of the scheduling method according to an embodiment of the present invention will be described. This scheduling method is based on the present invention. In this scheduling method, when dynamic resource allocation is performed in a TDM network, recalculation is performed for all resource allocations every recalculation period such as a traffic fluctuation period. Instead of performing the calculation, recalculation considering only the traffic fluctuation during this period reduces the amount of calculation, and asymptotically approaches the optimal solution for time slot allocation at high speed. In this method, after calculating the required TS matrix, only the portion that has changed in comparison with the previously calculated required TS matrix is changed in the time slot allocation in the link schedule table. For example, the time slot allocation is changed by reducing the number of assigned time slots for the ground (path) in which the traffic volume has decreased and the required number of time slots has been reduced. This is performed by assigning an additional time slot to the ground where the traffic volume has increased (this is called a time slot pasting (Re-allocation) process).

  FIG. 8 shows the concept of the scheduling method of the present embodiment by showing changes in the link schedule table. In a unidirectional ring network having five nodes A to E similar to those shown in FIG. 1, it is assumed that the contents of the link schedule table at a certain point in time are as shown on the left side of FIG. Here, if there is a traffic fluctuation, the traffic volume of each ground of C → A, A → D and D → A is decreased, and the traffic of each ground of A → B and B → E is increased. In the scheduling method based on FIG. 8, as shown on the right side of FIG. 8, the allocated time slots are stripped for the respective grounds of C → A, A → D, and D → A, and the stripped portions are respectively A → B and B → E. Paste to the ground. As a result, bandwidth is interchanged among a plurality of paths. At this time, by calculating only for the ground where the required traffic volume has changed, the calculation volume is greatly reduced and the calculation speed is improved, and the bandwidth (ie, time slot) is dynamically retracked following the traffic fluctuation. A large scale network system that can be allocated is realized.

  FIG. 9 is a diagram schematically showing the scheduling method of the present embodiment. In the present embodiment, attention is paid to the ground (path) where the fluctuation of the requested traffic amount is large between the previous time and the current time, and the bandwidth used in the ground where the requested amount is greatly reduced is changed to the ground where the requested amount is greatly increased. assign. For example, as shown in FIG. 9A, it is assumed that, among the nodes A and B on the network, the amount of traffic for the node A is large and a larger band is allocated for the node A. In this state, as shown in FIG. 9B, when the traffic volume for node A increases, the traffic volume for node B increases, and the bandwidth for B is insufficient. become. Therefore, in the method of the present embodiment, the operation of cutting the remaining band for A (that is, the stripping process) and allocating the band for the remaining band for B (that is, the pasting process) is performed entirely. Execute for the ground. As shown in FIG. 9C, the calculation time for stripping or pasting increases according to the number of stripping and pasting executions (number of target paths), but the time slot allocation Efficiency is improved and the actual bandwidth utilization in the network can be improved. Therefore, it is sufficient that the inefficiency is significantly improved by a small number of stripping and pasting processes. For this purpose, as will be described later, a path for performing stripping and pasting processes is determined in any order, and a link is made. The algorithm on which position on the schedule table is to be stripped or pasted is important.

  Hereinafter, the scheduling method of this embodiment will be described in more detail. In the following description, a one-way ring network having five nodes A to E as shown in FIG. 1 is considered, and one channel can be connected to the same time slot at the same time per link (ie, fiber multiplexing or Wavelength multiplexing is not performed). For example, such a network is an optical network using a single wavelength, and has a time slot length of 100 μs and a frame length of 5 ms. However, the present invention can be applied to a network having an arbitrary topology including a bidirectional ring, and can also be applied to a network that performs wavelength multiplexing. When wavelength multiplexing is performed, a link schedule table is created on each wavelength of each fiber.

  FIG. 10 shows an example of partial rewriting of the link schedule table in the present embodiment.

  Assume that the contents of the link schedule table 25 at a certain point in time are as shown in FIG. Here, traffic fluctuation occurs until the next schedule trigger, at which time the number of requested time slots (TS) decreases by 1 for the ground C → A, decreases by 2 for B → E, decreases by 1 for A → D, It is assumed that C → D increases by 1, E → A increases by 1, C → E increases by 1, and A → E increases by 1. First, a stripping process is performed as shown in FIG. 10B, and as a result, the time slots indicated by the broken lines in the figure are reduced. Next, as shown in FIG. 10C, a pasting process for assigning the reduced bandwidth to the increased bandwidth is performed. The time slots increased (pasted) by the pasting process are hatched in the figure. Here, the stripping process is performed prior to the pasting process. However, as will be described later, the order of the stripping process and the pasting process may be reversed.

  Next, the stripping process will be described. In the stripping process, a time slot is deleted for a path whose required time slot has decreased due to traffic fluctuation. At this time, an object is to create a continuous free area as large as possible on the link schedule table. By creating a large continuous free area, it is easy to set long paths (paths that pass through more links) and thick paths (paths that combine multiple time slots to increase the bandwidth) in the pasting process. Become.

  In order to create a continuous free space as much as possible, it is important to determine the order of stripping which path is to be stripped first, and the position to strip which path slot is prioritized for a certain path. It is. FIG. 11 illustrates the determination of the order of stripping. In the figure, hatched time slots are time slots assigned to paths with no change in traffic volume, and are not subject to stripping processing. As shown in the upper side of the figure, it is possible to reduce 3 time slots (TS) in a state where 5 time slots are assigned to paths B → D, and 2 time slots are assigned to A → E. It is assumed that 1 time slot can be reduced and 1 time slot can be reduced in a state where 2 time slots are assigned to the paths D → E. At this time, since a path that occupies a large area on the link schedule table is a hindrance, it is preferentially stripped from a long path and a path with a large number of variable time slots. As a result, as indicated by the numbers [1] to [3] on the lower side of FIG. 11, the slots assigned to the paths may be reduced in the order of A → D, B → D, and D → E. It will be.

  FIG. 12 illustrates that when a plurality of time slots are assigned to the same path, a position to be preferentially stripped from among the time slots assigned to the path is described. Here, it is shown in which order the time slots are stripped from the path B → D to which the four time slots are assigned, and whether the time slots are stripped in the order from the path C → E to which the two time slots are assigned. Yes. Since we want to create a large empty area on the link schedule table, there are empty time slots in the surroundings (adjacent time slots on the link schedule table, that is, the same time slot position on the adjacent link and the previous and subsequent time slot positions on the same link). Priority is given to stripping from many time slots. As a result, as indicated by the numbers [1] to [4] on the lower side of FIG. 12, for example, for the path B → D, the time slots are deleted in the order of the time slot positions S3, S2, S4, and S1. It will be good.

  Next, the pasting process will be described. In the pasting process, a time slot is added for a path whose required time slot has increased due to traffic fluctuation. At this time, the purpose is to fill a gap without minimizing a continuous free area on the link schedule table as much as possible. . By making the continuous empty area as small as possible, it becomes easy to set a long path and a thick path in the subsequent pasting process.

  In order not to make the continuous free space as small as possible, the order of pasting which path is pasted first and the pasting position of which time slot position is pasted for a certain path Decision is important. FIG. 13 illustrates the determination of the order of pasting. In the figure, hatched time slots are already assigned time slots. Consider a case in which one time slot is pasted on path A → B, one time slot is pasted on A → E, and two time slots are pasted on B → D. At this time, a significant increase in traffic may lead to buffer overflow or loss, and considering that it is difficult to set a long path, it is pasted in preference to a path with a large fluctuation in the number of required time slots or a long path. To do. As a result, as shown by the numbers [1] to [3] on the lower side of FIG. 13, the paths need only be pasted in the order of B → D, A → E, and A → B.

  FIG. 14 illustrates which slot position should be preferentially pasted when a plurality of time slots are pasted for the same path. Here, regarding the path B → D, two divided time slots have already been assigned, and further time slots are assigned to the path B → D. Because we want to fill in the gap as much as possible on the link schedule table, there are few free time slots in the surroundings (adjacent time slots on the link schedule table, that is, the same time slot position on the adjacent link and the time slot positions before and after the same link) Pasting is performed with priority from the time slot. As a result, as indicated by the numbers [1] to [4] in the figure, time slots may be added with priority in the order of time slot positions S4, S6, S2, S8, and S1.

  In the present embodiment, the empty space information in the link schedule table is particularly managed to simplify the processing required to accommodate paths with increased traffic and newly registered paths. This management is performed by the table empty management unit 34 in the scheduler 20, as will be described later. FIG. 15 shows a process for accommodating a new path by managing empty information.

  A time slot indicated by a broken-line frame on the left side of FIG. 15 is a time slot that has become free due to the deallocation being reduced by stripping. A time slot indicated by a solid frame is a time slot that is currently used. In this embodiment, the free information in the link schedule table is managed, and the information on the path to be increased is held in the form of an increased path list, and the unassigned in the link schedule table based on the free information. The increased paths are allocated to the empty time slots.

  The scheduling method in the present embodiment described above is executed by the scheduler 20 provided in any one of the nodes 11 in the TDM network system shown in FIG. At this time, in the scheduler 20, as described below, a function of detecting a difference in traffic information between the previous time and the current time, a function of performing a stripping process, and a function of performing a pasting process are important. Although the scheduler 20 can be configured as dedicated hardware, a general-purpose computer having a microprocessor, a memory, a communication interface, and the like is used, and a computer program for executing the function of the scheduler 20 is executed on the computer. Can also be realized. When the scheduler 20 is realized by executing a program on a computer, the link schedule table is stored and stored in a memory constituting the computer.

  A host computer 12 connected to the node 11 can be of a known configuration. The node 11 needs to have a configuration in which a function of dynamically collecting a traffic amount of a path and notifying the scheduler 20 is added to a known configuration.

  FIG. 16 shows a typical configuration example of a node used in the optical TDM network. The node includes a switch 91 for switching the route of data transferred through the link in units of time slots, an interface (I / O) 92 for exchanging data with the host computer 12, and the host computer 12. A buffer 93 that temporarily stores data transmitted from the side, a transmission unit (Tx) 94 that receives transmission data from the buffer 93, generates an optical signal sent to the link, and sends it to the switch 91; A receiving unit (Rx) 95 that receives a signal and sends it as received data to the host computer 12 via the I / O 92 and a controller 96 that controls the entire node 11 are provided. In particular, the controller 96 controls the switch 91 and the transmission unit 94 based on the node schedule table (processing schedule) transmitted from the scheduler 20. The function of dynamically collecting the traffic amount of the path and notifying the scheduler 20 is, for example, any one of the switch 91, the interface 92, the buffer 93, the transmission unit 94, and the reception unit 95 among the blocks constituting the node 11. Can be provided.

  In the scheduling method of the present embodiment, there are various implementation methods depending on the order of performing the stripping process and the pasting process. A representative one of them is referred to as Method 1 to Method 3.

  In the method 1, the path where the traffic is reduced is stripped, and then the path where the traffic is increased is pasted. Method 1 performs band allocation according to traffic flow.

  In method 2, the pasting process is performed with priority, and the stripping process is performed as necessary (when there is no free space to be allocated). In this method, the surplus bandwidth can be effectively utilized.

  Method 3 is a combination of method 1 and method 2 described above, and method 1 and method 2 are selectively used depending on whether there are many paths with reduced traffic or many paths in the network.

  Further, as an example of a change common to the above-described method 1 to method 3, there is one that limits the number of executions of the stripping process and the pasting process. In this modified example, a good solution is obtained within a predetermined time by aborting the calculation in the middle so that it converges quickly to the optimal solution for time slot allocation.

  Hereinafter, the structure for implementing the method 1 to the method 3 and those in which changes are made will be described.

  First, method 1 will be described. FIG. 17 shows the configuration of the scheduler 20 used in the TDM network system that performs scheduling based on the method 1.

  The scheduler 20 periodically aggregates the AC traffic volume, detects a traffic fluctuation difference between each ground, and executes a process of partially changing the link schedule table 25. Further, by managing the free space in the link schedule table 25, the search for the entire link schedule table is not required, and the calculation cost is reduced. The scheduler 20 based on the method 1 includes an AC traffic amount totaling unit 21, a traffic variation difference calculating unit 31, a variation difference table 32, a decrease stripping calculation unit 33, a table empty management unit 34, and an increment pasting calculation unit. 35, an allocation history management unit 36, a link schedule table 25, a table conversion unit 26, a node schedule table 27, a table transmission unit 28, and a timer 29. Among them, the AC traffic volume totaling unit 21, the link schedule table 25, the table conversion unit 26, the node schedule table 27, the table transmission unit 28, and the timer 29 are the same as those provided in the conventional scheduler 20 shown in FIG. Is.

  The AC traffic volume totaling unit 21 obtains traffic information periodically reported from each node 11 in the network, calculates the required number of time slots between all the grounds in the network, and sends this to the traffic fluctuation difference calculating unit 31. hand over. The traffic fluctuation difference calculation unit 31 compares the number of requested time slots obtained from the AC traffic amount totaling unit 21 with the number of requested time slots obtained at the immediately preceding control timing, so that the traffic fluctuation up to the corresponding timing can be calculated. The difference is calculated and stored in the fluctuation difference table 32 as fluctuation difference information. As will be described later, the variation difference information written in the variation difference table 32 is subject to classification and sorting based on the variation amount. If necessary, the decrease difference calculation unit 33 and the increase paste calculation unit 35 are used. Passed to.

  The decrease stripping calculation unit 33 converts the fluctuation difference information stored in the fluctuation difference table 32, the table empty information obtained from the table empty management unit 34, and the past allocation history obtained from the allocation history management unit 36. Based on the result, the table empty information in the table empty management unit 34 and the link schedule table 25 are rewritten, and the allocation bandwidth obtained by the current peeling process is transferred to the allocation history management unit 36. Notice. Similarly, the increment pasting calculation unit 35 includes the variation difference information stored in the variation difference table 32, the table empty information obtained from the table empty management unit 34, and the past allocation history obtained from the allocation history management unit 36. Based on the result, the table empty information and the link schedule table 25 in the table empty management unit 34 are rewritten according to the result, and the allocation bandwidth by the pasting process is assigned to the allocation history management unit 36 is notified.

  The table empty management unit 34 holds and manages information on empty slots in the link schedule table 25. The table empty management unit 34 receives a rewrite notification from the decrease stripping calculation unit 33 and the increase paste calculation unit 35, and also creates a link schedule. The table 25 is notified of vacant information in the table, and provides the vacant information to the reduced-stripping calculation unit 33 and the increased-pasting calculation unit 35 prior to the execution of the stripping process or the pasting process. The allocation history management unit 36 holds and manages the allocation history. The allocation history management unit 36 receives an allocation band notification from the decrease stripping calculation unit 33 and the increase paste calculation unit 35 and adds it to the allocation history. Allocation history information is provided to the stripping calculation unit 33 and the increment paste calculation unit 35.

  The link schedule table 25 is a table representing the vacancy status of the time slots defined for each link in the network, and is a stripping process and pasting by the decrease stripping calculation unit 33 and the increment paste calculation unit 35. Rewritten by processing. The link schedule table 25 is configured to provide the empty information to the empty table manager 34.

  The table conversion unit 26 converts the link schedule table 25 into a node schedule table 27 for each node 11, and the table transmission unit 28 assigns the processing schedule of the node to each node 11 based on the node schedule table 27. Send regularly. The node schedule table 27 itself may be transmitted to each node 11 instead of the processing schedule. The timer 29 is for synchronizing with each node 11 and performing time management. In particular, by outputting a trigger to the fluctuation difference table 32 and the allocation history management unit 36, the timer 29 is periodically updated. Initialization (clearing operation) of the fluctuation difference table 32 and the allocation history management unit 36 can also be executed.

  Next, the operation of the scheduler 20 will be described using the sequence diagrams shown in FIGS.

  First, as shown in FIG. 18, information on the traffic volume is sent from each node to the AC traffic volume totaling unit 21 (step 101). In step 102, the AC traffic volume totaling unit 21 sets the required time slot between nodes, that is, each ground. The number of (TS) is calculated and sent to the traffic fluctuation calculation unit 31 as a TS matrix (step 103). In step 104, the traffic fluctuation difference calculation unit 31 compares the TS matrix sent from the AC traffic amount totaling unit 21 with the TS matrix acquired last time, and calculates the difference (variation difference information) in the requested time slot number. The data is sent to the table 32 (step 105). In step 106, the fluctuation difference information is written into the fluctuation difference table 32. From the fluctuation difference table 32, information on the path with the reduced number of required time slots and the amount of the decrease are sent to the decrementing calculation unit 33 (step 107), and the path with the increased number of required time slots and the amount of the increase. Is sent to the increment paste calculator 35 (step 108). Also, the bandwidth allocation history of each path is sent from the allocation history management unit 36 to the decrease stripping calculation unit 33 and the increase paste calculation unit 35 (steps 109 and 110).

  In step 111, the reduction stripping calculation unit 33 calculates the stripping process, and based on the calculation result, sends the allocation history to the allocation history management unit 36, and uses an empty update trigger to update the table free information. Is sent to the table empty management unit 34 (step 112), and a table update trigger is sent to the link schedule table 25 to update the link schedule table 25 (step 113). Accordingly, the table empty management unit 34 updates the table empty information in step 114, and updates the link schedule table 25 in step 115.

  Next, from the fluctuation difference table 32, information on the path with the increased number of required time slots and the amount of the increase are sent to the increment pasting calculation unit 35 (step 116), and the bandwidth of each path from the allocation history management unit 36. The allocation history is sent to the increment paste calculation unit 35 (step 117). In response to this, the increment paste calculation unit 35 calculates paste processing in step 118, and based on the calculation result, sends the allocation history to the allocation history management unit 36 and updates the free information in the table. An empty update trigger is sent to the table empty manager 34 (step 119), and a table update trigger is sent to the link schedule table 25 for updating the link schedule table 25 (step 120). Accordingly, the table empty management unit 34 updates the table empty information in step 121, and updates the link schedule table 25 in step 122.

  After that, as shown in FIG. 19, when the table update in the link schedule table 25 is completed in step 131, the contents of the link schedule table 25 are sent to the table conversion unit 26 (step 132), and the table conversion unit 26 performs step 133. , The processing schedule for each node is derived and the derivation result is sent to the node schedule table 27 (step 134), and in step 135, the derivation result is stored as a node schedule table. Thereafter, the contents of the node schedule table 27 are sent to the table transmission unit 28 (step 136), and the table transmission unit 28 executes distribution processing for each node 11 in step 137. As a result, the processing schedule or node schedule table for each node is sent to each node (step 138).

  Thus, the scheduling by the method 1 is completed. Of the above processing, the processing in steps 101 and 102 and the processing after step 131 (processing shown in FIG. 19) are the same as those in the conventional scheduler shown in FIG.

  Next, the scheduler 20 for executing the method 1 will be described in more detail with a specific example.

  The AC traffic volume totaling unit 21 periodically collects all ground-to-ground traffic information from each node and manages the required number of time slots in the form of a matrix. The node A to each of the nodes B to E Assume that the traffic band is as shown in FIG. Then, the AC traffic volume totaling unit 21 converts these traffic bands, for example, at a rate of one time slot per 10 Mbps (see (b) of FIG. 20), and the result is obtained as shown in (c) of FIG. Store as a required TS matrix in memory. This required TS matrix is sent to the traffic fluctuation difference calculation unit 31.

  The traffic fluctuation difference calculation unit 31 holds the required TS matrix sent last time from the AC traffic amount totaling unit 21 in its memory, and as shown in FIG. 21, the previous time from the required TS matrix sent this time. An operation for comparing and subtracting the required TS matrix sent for each matrix element is performed to calculate a two-dimensional array representing the varied difference, and this two-dimensional array is stored in the variation difference table 32. The fluctuation difference table 32 is managed in the memory of the scheduler 20, for example.

  Next, the stripping (Rip-up) processing and pasting (Re-allocation) processing in Method 1 will be described. In Method 1, after the stripping process is performed on the ground with a decreased traffic volume, the pasting process is performed on the ground with an increased traffic volume. As described above, each process acquires information necessary for time slot allocation from the fluctuation difference table 32, the table empty management unit 34, and the allocation history management unit 36, and then the link schedule table 25 and the table empty management unit 34. This is a process for changing the state. Therefore, among the functional blocks constituting the scheduler 20, the variation difference table 32, the decrease stripping calculation unit 33, the table empty management unit 34, the allocation history management unit 36, and the link schedule table 25 are directly related to the stripping process. The variation difference table 32, the increment pasting calculation unit 35, the table empty management unit 34, the allocation history management unit 36, and the link schedule table 25 are directly related to the pasting process.

  First, the stripping process in Method 1 will be described in detail. In the stripping process, the stripped-off calculating unit 33 obtains free information in the link schedule table 25 with reference to the table empty managing unit 34 for the ground whose traffic difference is indicated in the fluctuation difference table 32. Then, the past bandwidth allocation information is acquired from the allocation history management unit 36, and the link schedule table is partially changed by reducing the number of time slots already allocated. This process includes a step of determining a ground selection order in which time slots are reduced, a step of determining the number of timeslots to be reduced, and a step of determining positions to reduce the time slots.

In the step of determining the ground selection order for reducing time slots, the stripping-off calculation unit 33 determines the ground selection order for stripping based on information from the variation difference table 32 and the allocation history management unit 36. To do. When the stripping process is performed, it is preferable that a continuous empty area as large as possible is formed on the link schedule table 25. Therefore, a long path and a path with a large fluctuation amount are preferentially reduced. In the example illustrated in FIG. 22, the cost value for each ground is calculated from the fluctuation difference in the number of time slots, and sorting is performed in descending order of the cost value, thereby determining the stripping execution order. In FIG. 22, the fluctuation amount a is the absolute value of the fluctuation difference in the number of time slots, the path length b indicates the number of links through which the path passes, and the random number c is an integer randomly generated between 1 and 99. , As cost value
Cost value = 10000 × a + 100 × b + c
Is used.

In the stage of determining the number of time slots to be reduced, the reduction amount stripping calculator 33 determines the number of time slots to be stripped from each ground. For example, the number of time slots corresponding to the amount of change is simply stripped off. Thereafter, in the stage of determining the position to reduce the time slot, the reduction amount stripping calculation unit 33 determines the time slot position to be stripped for each ground to be stripped based on the information from the table empty management unit 34. decide. FIG. 23 shows how to determine the position of the time slot to be stripped. In the link schedule table portion on the left side of the figure, the hatched time slot is a time slot in use. When determining the time slot position to be stripped off, the cost value is calculated from the availability in the link schedule table 25, and the position is selected in descending order of the cost value. When stripping processing is performed, it is preferable that a continuous free area as large as possible is formed on the link schedule table 25, so that adjacent links are empty at the same time slot position or the same link. In the case where the adjacent time slots are empty, priority is given to the time slot reduction target. In the illustrated example, the same TS adjacent sky α indicates the number of links that are vacant among the adjacent links at the same time slot position, and the same link adjacent sky β indicates the time slot adjacent in the same link. It shows the number of empty time slots. For the time slot position γ, the time slot number assigned from 1 is used as it is. Here, as the cost value,
Cost value = 10000 × α + 100 × β + γ
Is used.

  In the example shown in FIG. 23, when one time slot is reduced from the three time slots (S3, S4, and S5) assigned to the ground B → C, S5 is set so that the continuous free space becomes as large as possible. The time slot indicated by is reduced.

  When the time slot to be reduced is determined as described above, the decrease stripping calculation unit 33 next reduces the time slot determined to be reduced from the link schedule table 25 and manages the table empty. The vacancy information held in the unit 34 is updated, the allocated bandwidth is notified to the allocation history management unit 36, and the allocation history management unit 36 is updated. The upper side of FIG. 24 shows time slot deletion in the link schedule table 25, and the deleted time slot is also registered in the table empty management unit 34 as empty. Since the time slots are reduced, the cost values of the surrounding time slots, specifically, the same TS adjacent sky α and the same link adjacent space β also change. 24 shows an example of allocation history information held in the allocation history management unit 36. The allocation history information is a record of how many timeslots are requested and how many timeslots are actually allocated each time scheduling is performed for each ground (path). Is stored in memory.

  In accordance with the order determined in the step of determining the ground selection order to reduce the time slots, the number of timeslots to be deleted and the position to be deleted are determined, the link schedule table 25 is changed, the table empty management unit 34 and the allocation history management unit 26 is repeated to complete the stripping process.

  Next, the pasting process in Method 1 will be described in detail. In the pasting process, the increment pasting calculation unit 35 acquires the free information in the link schedule table 25 with reference to the table empty management unit 34 for the ground where the traffic difference is indicated in the fluctuation difference table 32. Then, the past bandwidth allocation information is acquired from the allocation history management unit 36, and the link schedule table is partially changed by increasing the allocated time slots. This process includes a step of determining a ground selection order for increasing time slots, a step of determining the number of timeslots to be increased, and a step of determining a position where the time slot is to be pasted.

At the stage of determining the selection order of the ground to increase the time slot, the increase paste calculation unit 35 determines the selection order of the ground to be pasted based on the information from the variation difference table 32 and the allocation history management unit 36. To do. When performing the pasting process, as described above, a time slot corresponding to a long path and a path with a large fluctuation amount is preferentially pasted. In the example shown in FIG. 25, the cost value for each ground is calculated from the difference in the number of time slots, and the order of stripping is determined by sorting in descending order of the cost value. The fluctuation amount a, the path length b, and the random number c are the same as those shown in FIG. As a cost value,
Cost value = 10000 × a + 100 × b + c
Is used.

In the stage of determining the increase number of time slots, the increase amount pasting calculation unit 35 determines the number of time slots to be pasted to each ground. For example, the number of time slots corresponding to the amount of variation is simply pasted. After that, at the stage of determining the position where the time slot is pasted, the increase amount pasting calculation unit 35 determines the time slot position for pasting for each ground to be pasted based on the information from the table empty management unit 34. decide. FIG. 26 shows how to determine the time slot position to be pasted. In the link schedule table portion on the left side of the figure, the hatched time slot is a time slot in use. When determining the time slot position to be pasted, the cost value is calculated from the availability in the link schedule table 25, and the position is selected in ascending order of the cost value. When performing the pasting process, it is preferable to fill in the gap as much as possible on the link schedule table 25, so that an adjacent link is in use at the same time slot position, or an adjacent time slot at the same link The ones that are in use are given priority for adding time slots. In FIG. 26, the same TS adjacent space α, the same link adjacent space β, and the time slot position γ are the same as those in FIG. Here, as the cost value,
Cost value = 10000 × (α + β) + γ
Is used.

  In the example shown in FIG. 26, when one more time slot is added to the ground D → E, there are S1, S2, S16, S18, S19, and S20 as candidate positions to be added. S16 having the smallest value is selected, and a time slot is assigned to this position.

  When it is determined which time slot to allocate as described above, the increment paste calculation unit 35 then allocates and adds the time slot determined to be pasted in the link schedule table 25, The vacancy information held in the table empty management unit 34 is updated, the allocated bandwidth is notified to the allocation history management unit 36, and the allocation history management unit 36 is updated. The upper side of FIG. 27 shows time slot assignment addition in the link schedule table 25, and the added time slot is deleted from the table empty management unit 34 as being no longer empty. Note that the addition of time slots also changes the cost values of surrounding time slots, specifically, the same TS adjacent sky α and the same link adjacent space β. The lower side of FIG. 27 shows an example of allocation history information held in the allocation history management unit 36, similar to that shown in FIG.

  In accordance with the order determined in the step of determining the ground selection order for increasing the time slots, determination of the number and position of addition of time slots, change of the link schedule table 25, table empty management unit 34 and allocation history management unit 26 is repeated, the pasting process is completed.

  An example of the stripping process and the pasting process will be described using equations.

  The number of nodes is N, and the number of time slots in one frame is Cap. When the time slot repetition period t is 5 ms and the time slot length TS is 100 μs, cap = 50. A pair (i.e., ground) of the transmission source node i and the destination node j is represented by (i, j). Of course, i ≠ j.

The number of hops in the path connecting (i, j) is expressed as Hop (i, j). For a ring network with N nodes, the minimum hop count is 1, the maximum hop count is N-1, and the average hop count is N / 2. A link connecting the paths between (i, j) is denoted by L _{i, j} (h), _and a link connecting (i, j) is denoted by k _{i, j} ( k),
The requested time slot number of the ground (i, j) at the scheduling opportunity T is set as Req _{i, j} (T), and the number of time slots assigned to the ground (i, j) at T is set as As _{i, j} (T). When performing semi-fixed scheduling, the scheduler may determine allocation at the next scheduling opportunity (T + 1) using information (Req _{i, j} (T), etc.) obtained at the opportunity T.

Let Hist _{i, j} (T) be the difference between the requested time slot number and the assigned time slot number in the allocation history over the past M times.

It is. As shown below, there is also a method using a difference from the _fair band C _fair (T).

The content of the link schedule table at opportunity T is expressed as τ (T, L _{i, j} (h), w, s) (h = 1,..., Hop (i, j))
The time slot number is s, and at time T, the link schedule table indicates that the link is L, and that the time slot is vacant at the time slot position s is represented by V (T, L, s) = 1.

In an example of the criteria for determining whether to perform stripping,
[1] From (i, j) that satisfies Req _{i, j} (T) <Req _{i, j} (T−1), time slots are set by Req _{i, j} (T−1) −Req _{i, j} (T). Strip or
[2] Req _{i, j} (T) ≦ As _{i, j} (T−1) <Req _{i, j From} ( _{i, j} ) that satisfies (T−1), As _{i, j} (T−1) −Req Strip time slots by _{i, j} (T).

An example of how to determine the order of stripping is to use the cost function Rip-up-priority
Rip-up-priority = μ {As _{i, j} (T−1) −Req _{i, j} (T)} + ν / Hop (i, j)
And strip off in descending order of the Rip-up-priority value. At this time, when μ> ν, a path with a large traffic fluctuation is selected with priority, and when μ <ν, a long path has a higher priority.

  In an example of a method for determining a time slot position to be stripped, a cost function Rip-up-gain for giving priority to a time slot having a lot of surroundings is assumed as stripping from the ground (i, j).

It is determined. This is Rip-up-gain
Rig-up-gain = α x [same TS adjacent sky] + β x [same slot adjacent sky]
Is equivalent to Then, the time slot position s that maximizes Rip-up-gain is selected. When a plurality of time slot positions have the same value, an old time slot position is selected from among them. Here, if α> β, the effect of easily setting a long path is strong, and if α <β, the effect of easily setting a thick path is strong.

In an example of criteria for determining whether or not to paste,
[1] Req _{i, j} (T)> Req _{i, j} (T−1) satisfying (i, j), time slots are set by Req _{i, j} (T) −Req _{i, j} (T−1). Paste or
[2] Req _{i, j} (T)> As _{i, j} (T−1) satisfying (i, j), the time slot is req _{i, j} (T) −As _{i, j} (T−1). paste.

An example of how to determine the order of pasting is to use the cost function Alloc-priority
Alloc-priority = δ {Req _{i, j} (T) −Req _{i, j} (T−1)} + ε · Hop (i, j)
And paste in descending order of Alloc-priority values. At this time, when δ> ε, a path with a large traffic fluctuation is selected with priority, and when δ <ε, a long path has a higher priority.

In an example of a method for determining the time slot position to be pasted, assuming that a time slot is pasted to the ground (i, j), it is first determined whether or not there is a band that can be pasted. This is h = 1,..., Hop (i, j), and for all h,
V (T, L _{i, j} (h), s) = 1
Is determined by whether or not Next, the cost function Alloc-cost to fill the gap as much as possible is

It is determined. Then, the time slot position s that minimizes Alloc-cost is selected. When a plurality of time slot positions have the same value, a young time slot position is selected from among them. Here, if γ> ω, the effect of easily setting a long path is strong, and if γ <ω, the effect of easily setting a thick path is strong.

  By the way, in the scheduling method based on the present invention, there are various variations as an algorithm for performing the stripping process and an algorithm for performing the pasting process, and different effects depend on how the cost function is set. Is obtained. Hereinafter, these variations will be described.

  Table 1 shows variations of algorithms for determining the selection order of paths to be stripped.

  Among the above, algorithms [2] to [4] are used in the above-described example in combination. It is also possible to use a combination of algorithms [1] to [9]. In addition, for sorting, classification based on threshold values may be used instead of sorting. In the classification based on the threshold, when one threshold is used (for example, when the path length is longer or shorter than the average value), when the path length exceeds the threshold, for example, the Randam Fit algorithm is used. Otherwise, for example, it may be possible to use the First Fit algorithm. A plurality of threshold values may be used, and different algorithms may be used for each class classified by the threshold values. The method using the threshold is intended to obtain the same effect as that by sorting while shortening the calculation time.

  Table 2 shows variations of the method for determining the time slot position to be stripped.

Algorithm [3] (determined in consideration of surrounding free space) is also used in the above-described example, but uses a cost function in order to consider the free space in the surroundings. The idea of how to build a cost function is to “create a long space” to make it easier to set a long path, “create a thick space” to make it easier to set a thick path, and a long and thick path. To make it easier to set, there is a "create long and thick path". These ideas can also be combined. In the following description, the count value is reflected in the cost function so that the time slot is more easily peeled off as the count value (the number of surrounding empty time slots) is larger.

  In order to “create a long space”, the number of surrounding spaces at the same time slot position may be counted. In the link schedule table shown in FIG. 28, considering the case where stripping is performed for three time slots assigned to the ground B → D, empty locations (in the figure, adjacent to the B → D path). It suffices to count the same time slot positions (indicated by ◎). At this time, among the time slot positions S2 to S4 of the B → D path, the count value for S2 is 1, the count value for S3 is 2, and the count value for S4 is 1. Further, the counting may be performed including a continuous empty portion (a portion indicated by ○ in the drawing) in addition to a portion indicated by “◎” so that a longer space is generated. Then, in the time slot of the B → D path, the count value of S2 becomes 2, and the count value of S3 becomes 3.

  In order to “create a thick space”, the number of surrounding free time slots on the same link is counted. In the example shown in FIG. 28, considering the case where stripping is performed from two time slots assigned to the ground C → E, there is no empty position in the same link at the time slot position S7. At slot position S8, as shown by ◎ in the figure, adjacent time slot position S9 is vacant in two links C to E, and the count value is 2. Further, as in the case described above, counting may be performed including a vacant place (time slot position S10 shown in FIG. 6) adjacent to the time slot position S9 so that a thicker vacant space is generated. In that case, the count value of the time slot position S8 is 4.

  In order to “create a long and thick space”, the surrounding space is counted. Not only empty positions adjacent to the same time slot position and the same link (indicated by ◎ in FIG. 28), but also locations adjacent to two of these ◎ (indicated by Δ in the figure) are counted. In the example shown in FIG. 28, the count value for the time slot position S8 is 4 for the path C → E.

  Table 3 shows variations of algorithms for determining the selection order of paths to be pasted.

  Among the above, algorithms [2] to [4] are used in the above-described example in combination. It is also possible to use a combination of algorithms [1] to [9]. Further, as in the case of the stripping process, classification based on threshold values may be used instead of sorting.

  Table 4 shows variations of the method for determining the time slot position to be pasted.

  Algorithm [3] (determined in consideration of surrounding free space) is also used in the above-described example, but uses a cost function in order to consider the free space in the surroundings. The idea of how to build a cost function is to “fill a short space” to make it easier to set a long path, “fill a thin space” to make it easier to set a thick path, and a long and thick path. To make it easier to set, there is a “fill short and narrow path”. These ideas can also be combined. In the following description, the count value is reflected in the cost function so that the time slot is more easily pasted as the count value is larger.

  In order to “fill a short space”, the number of assigned time slots at the same time slot position may be counted. In the link schedule table shown in FIG. 29, hatched portions indicate already assigned time slots. Here, considering that a time slot of ground B → D is newly allocated, the candidate time slot positions are, for example, S2, S3, and S4. In this case, it is only necessary to count the assigned locations (locations indicated by に お い て in the figure at the same time slot position) adjacent to the B → D path. At this time, among the time slot positions S2 to S4 of the B → D path, the count value for S2 is 1, the count value for S3 is 0, and the count value for S4 is 1. Furthermore, the counting may be performed including the allocated locations (locations indicated by ○ in the figure) that are continuous to the adjacent links at the same time slot position. Then, the count value of S4 becomes 2 in the time slot of the path of B → D.

  In order to “fill a narrow space”, the number of assigned time slots around the same link is counted. In the example in which the pasting process is performed in the path B → D shown in FIG. 29, there is no allocated time slot position adjacent in the same link at the time slot position S3, whereas the time slot position S2 shows As shown in FIG. 2, the adjacent time slot positions S1 have already been assigned in two links B to D, and the count value is 2. Similarly, the count value at the time slot position S4 is 1. Further, similarly to the case described above, the counting may be performed including the already assigned locations (location of the time slot position S6 indicated by ○ in the figure) that are further continuous in adjacent positions in the same link. In that case, the count value of the time slot position S4 is 2.

  In order to “fill a short and narrow space”, surrounding assigned time slots are counted. Not only assigned locations adjacent to the same time slot position or the same link (locations indicated by ◎ in FIG. 29) but also locations adjacent to two of these ◎ locations (locations indicated by Δ) are counted. In the example shown in FIG. 29, regarding the path B → D, the count value for the time slot position S2 is 4, and the count value for the time slot position S4 is 3.

  Next, method 2 will be described. In Method 2, the difference between the previous actual bandwidth allocation and the current required TS matrix is not determined by comparing the previous and current required TS matrixes and the difference between the actual bandwidth allocation and the current required TS matrix is determined as “allocation difference”. It is defined and used for the calculation of the stripping process and the pasting process. Also, the pasting process based on the shortage in the divergence (allocation difference) is performed with priority, and the stripping process is performed based on the excess in the allocation difference only when necessary. FIG. 30 shows the configuration of the scheduler 20 used in the TDM network system that performs scheduling based on the method 2.

  The scheduler 20 used in the method 2 and shown in FIG. 30 is the same as the scheduler 20 for the method 1 shown in FIG. 17, except that a traffic fluctuation difference calculation unit, a fluctuation difference table, a decrease stripping calculation unit, and An increase paste calculation unit is not provided, but instead an allocation difference calculation unit 41, an allocation difference table 42, an excess stripping calculation unit 43, and an insufficient paste calculation unit 44 are provided. It is different from that shown in FIG. Hereinafter, Method 2 will be described focusing on this difference.

  The allocation difference calculation unit 41 calculates the allocation difference by comparing the number of time slots actually allocated at the previous control timing (that is, the previous time) with the requested time slot number obtained from the AC traffic amount totaling unit 21. And stored in the allocation difference table 42 as allocation difference information. Information on the number of lime slots actually allocated last time is supplied from the allocation history management unit 36. The allocation difference information written in the allocation difference table 42 is subject to classification and sorting in the same manner as described above, and is passed to the excess stripping calculation unit 43 and the shortage pasting calculation unit 44 as necessary.

  The excess stripping calculation unit 43 is equivalent to the reduction stripping calculation unit, and includes the allocation difference information stored in the allocation difference table 42 and the table empty information obtained from the table empty management unit 34 and the allocation. A stripping process is executed based on the past allocation history obtained from the history management unit 36, and the table empty information in the table empty management unit 34 and the link schedule table 25 are rewritten according to the result. The allocation history management unit 36 is notified of the allocated bandwidth by the stripping process. Similarly, the shortage paste calculation unit 46 is equivalent to the increase paste calculation unit, and includes allocation difference information stored in the allocation difference table 42 and table empty information obtained from the table empty management unit 34. And past allocation history obtained from the allocation history management unit 36, pasting processing is executed, and table empty information in the table empty management unit 34 and the link schedule table 25 are rewritten according to the result, Then, the allocation history management unit 36 is notified of the allocated bandwidth by the pasting process.

  The allocation history management unit 36 holds and manages the allocation history, and receives the notification of the allocated bandwidth from the excess stripping calculation unit 43 and the shortage pasting calculation unit 44 and adds it to the allocation history. Allocation history information is provided to the stripping calculation unit 43, the shortage pasting calculation unit 44, and the allocation difference calculation unit 41.

  Next, the operation of the scheduler 20 will be described with reference to sequence diagrams shown in FIG. 31 and FIG.

  First, information on the traffic volume is sent from each node to the AC traffic volume totaling unit 21 (step 201). In step 202, the AC traffic volume totaling unit 21 calculates the number of time slots (TS) required between the nodes, that is, each ground. Then, it is sent to the allocation difference calculation unit 41 as a TS matrix (step 204). Prior to this, the allocation history management unit 36 sends the previously allocated TS matrix information to the allocation difference calculation unit 41 (step 203).

  In step 205, the allocation difference calculation unit 41 compares the TS matrix sent from the AC traffic volume totaling unit 21 with the TS matrix of the previous allocation, and sends the difference (allocation difference information) to the allocation difference table 42. (Step 206) In step 207, allocation difference information is written into the allocation difference table. From the allocation difference table 42, information on the path in which the number of allocated time slots exceeds the required number of time slots and the excess amount is sent to the excess stripping calculation unit 43 (step 208), and the allocated time Information regarding the path whose number of slots is insufficient compared to the number of required time slots and the amount of the shortage is sent to the shortage pasting calculation unit 44 (step 209). Further, the bandwidth allocation history of each path is sent from the allocation history management unit 36 to the excess stripping calculation unit 43 and the shortage pasting calculation unit 44 (steps 210 and 211).

  In step 212, the insufficient portion pasting calculation unit 44 calculates pasting processing. Next, in step 217, it is determined whether pasting is impossible in the calculation in step 212, and pasting is impossible. If so, a stripping command is sent to the stripping calculation unit 43 (step 218). In step 219, the excess stripping calculation unit 43 determines whether a stripping command has been received. If it has been received, the excess stripping calculation unit 43 calculates the stripping process in step 220, and based on the calculation result, sends the allocation history to the allocation history management unit 36 and updates the free information in the table. Therefore, an empty update trigger is sent to the table empty manager 34 (step 221), and a table update trigger is sent to the link schedule table 25 for updating the link schedule table 25 (step 222). Accordingly, the table empty management unit 34 updates the table empty information in step 223, and updates the link schedule table 25 in step 224. Further, the excess stripping calculation unit 43 sends a stripping completion notification to the shortage pasting calculation unit 44 (step 225).

  In step 226, the shortage paste calculation unit 44 determines whether or not a stripping completion notification has been received. If received, the shortage paste calculation unit 44 performs paste processing calculation in step 227 and assigns the allocation based on the calculation result. The history is sent to the allocation history management unit 36, and an empty update trigger is sent to the table empty management unit 34 to update the table empty information (step 228), and the table update trigger is updated to update the link schedule table 25. Is sent to the link schedule table 25 (step 229). Accordingly, the table empty management unit 34 updates the table empty information in step 230, and updates the link schedule table 25 in step 231.

  Thereafter, the processing from step 131 in method 1 (see FIG. 19) is executed.

  Next, method 3 will be described. In method 3, method 1 and method 2 are combined. When there is little ground with increased traffic, method 2 is applied to give priority to the pasting process, and when there is much ground with increased traffic, method 1 is used. Apply and give priority to the stripping process. FIG. 33 shows the configuration of the scheduler 20 used in the TDM network system that performs scheduling based on the method 3.

  The scheduler 20 used in the method 3 and shown in FIG. 33 is further provided with an allocation difference calculation unit 41 and an allocation difference table 42 for implementing the method 2 in the scheduler 20 for the method 1 shown in FIG. A comparator 51 is provided for determining which of the process 1 and method 2 is prioritized. Further, instead of the decrease peeling calculation unit, a peeling calculation unit 52 having the functions of the decrease peeling calculation unit and the excess peeling calculation unit in the method 2 is provided, and the increase paste calculation unit is provided. Instead of this, a pasting calculation unit 53 having the functions of the increment pasting calculation unit and the shortage pasting calculation unit in method 2 is provided. The comparator 51 is supplied with the fluctuation difference information from the fluctuation difference table 31, compares the number of grounds with increased traffic and the number of grounds with decreased traffic, and depending on which is larger, the stripping calculation unit 52 and the pasting calculation. The unit 53 is instructed to perform the stripping process and the pasting process by either the method 1 or the method 2.

  Next, the operation of the scheduler 20 will be described with reference to the sequence diagrams shown in FIGS.

  First, information on the traffic volume is sent from each node to the AC traffic volume totaling unit 21 (step 301). In step 302, the AC traffic volume totaling unit 21 calculates the number of required time slots (TS) between the nodes, that is, each ground. Then, it is sent as a TS matrix to the traffic fluctuation difference calculation unit 31 and the allocation difference calculation unit 41 (steps 304 and 305). Prior to these, the allocation history management unit 36 sends the previously allocated TS matrix information to the allocation difference calculation unit 41 (step 303).

  In step 306, the traffic fluctuation difference calculation unit 31 compares the required TS matrix between the previous time and the current time, sends the fluctuation difference information to the fluctuation difference table 32 (step 308), and the fluctuation difference information fluctuates in step 310. It is written in the difference table 32. In parallel with this, in step 307, the allocation difference calculation unit 41 compares the current required TS matrix with the TS matrix of the previous allocation, and sends those differences (allocation difference information) to the allocation difference table 42 (step 309), the allocation difference information is written in the allocation difference table 42 in step 311. The bandwidth allocation history of each path is sent from the allocation history management unit 36 to the stripping calculation unit 52 and the pasting calculation unit 53 (steps 312 and 313).

  Next, from the fluctuation difference table 32, information on the path with the reduced number of required time slots and the amount of decrease is sent to the stripping calculation unit 52 (step 314), and the path with the increased number of required time slots and the amount of increase thereof. Is sent to the increment paste calculation unit 53 (step 315). From the allocation difference table 42, information on the path in which the number of allocated time slots exceeds the required number of time slots and the excess amount is sent to the excess stripping calculation unit 43 (step 316). Information regarding the path whose number of time slots is insufficient compared to the required number of time slots and the amount of the shortage is sent to the shortage pasting calculation unit 44 (step 317). Further, information on the number of grounds where the traffic has decreased and the number of grounds where the traffic has increased is sent from the fluctuation difference table 32 to the comparator 51 (step 318).

  In step 319, the comparator 51 compares the number of grounds with decreased traffic with the number of grounds with increased traffic. In step 320, the comparator 51 determines which number of grounds is larger. Here, if the increased number of grounds ≧ the decreased number of grounds, the comparator 51 instructs the stripping calculating unit 52 and the pasting calculating unit 53 to execute the method 1 (steps 321 and 322) and increases. If the number of grounds is smaller than the number of grounds, the execution of method 2 is instructed to the stripping calculation unit 52 and the pasting calculation unit 53 (steps 323 and 324). In response to this, the stripping calculation unit 52 and the pasting calculation unit 53 perform the scheduling calculation by the instructed method (method 1 or method 2) in step 325, and based on the calculation results, the allocation history is calculated. In addition to sending to the allocation history management unit 36, an empty update trigger is sent to the table empty management unit 34 for updating the table empty information (steps 326, 327), and a table update trigger is used to update the link schedule table 25. Is sent to the link schedule table 25 (steps 328 and 329). As a result, the table empty management unit 34 updates the table empty information in step 330, and updates the link schedule table 25 in step 331.

  Thereafter, the processing from step 131 in method 1 (see FIG. 19) is executed.

  Next, an example in which the method 1 to the method 3 are further changed will be described.

  As described above, the number of executions of the stripping process and the pasting process can be limited during scheduling. Here, the case where the number of executions of the stripping process and the pasting process is limited when the method 1 is performed will be described, but the number of executions can be similarly limited in the method 2 and the method 3.

  FIG. 36 shows a configuration of the scheduler according to the method 1 and capable of limiting the number of executions of the stripping process and the pasting process. The scheduler 20 shown in FIG. 36 has a bandwidth for managing the number of times the bandwidth has been changed to further limit the number of executions of the stripping process and the pasting process to the scheduler for the method 1 shown in FIG. A change counter 37 is provided. The band change counter 36 is provided with a counter that counts the number of times the stripping process has been performed and a counter that counts the number of times that the pasting process has been performed. An upper limit Ra of the number of executions is set. The upper limit values Ru and Ra can be determined independently in advance.

  FIG. 37 is a sequence diagram showing a scheduling process using the scheduler 20 shown in FIG. Here, processing after the processing in which the difference in the number of requested time slots is sent from the traffic fluctuation difference calculation unit 31 to the fluctuation difference table 32 will be described. The processing before this is the same as that shown in FIG.

  When the traffic fluctuation difference calculation unit 31 sends the difference in the number of requested time slots (fluctuation difference information) to the fluctuation difference table 32 (step 105), the fluctuation difference information is written into the fluctuation difference table 32 in step 106. From the fluctuation difference table 32, information on the path with the reduced number of required time slots and the amount of the decrease are sent to the decrementing calculation unit 33 (step 107), and the path with the increased number of required time slots and the amount of the increase. Is sent to the increment paste calculator 35 (step 108). The bandwidth allocation history of each path is sent from the allocation history management unit 36 to the decrease stripping calculation unit 33 and the increase paste calculation unit 35 (steps 108 and 109).

  Then, in step 151, the reduction stripping calculation unit 33 calculates a stripping process for one path (ground), and based on the calculation result, sends an allocation history to the allocation history management unit 36 and An update trigger is sent to the table empty management unit 34 (step 152), a notification that stripping processing has been performed (peeling processing notification) is sent to the band change counter 37 (step 153), and the table update trigger is sent to the link schedule table. 25 (step 154). Accordingly, the table empty management unit 34 updates the table empty information in step 155, and updates the link schedule table 25 in step 156. In the band change counter 37, the stripping number counter is incremented in step 157, and it is determined in step 158 whether or not the counter value has reached the upper limit value Ru. If the upper limit value Ru has not been reached, the band change counter 37 returns to the state of waiting for the next stripping process notification. If the upper limit value Ru has been reached, the bandwidth change counter 37 sends an abort instruction to the decrement stripping calculation unit 33 (step). 159). The decrement removal calculation unit 33 determines in step 160 whether or not an abort instruction has been issued. If there is no abort instruction, the process returns to step 151 to execute a calculation for the next one-pass exfoliation process. To do.

  On the other hand, the increment paste calculation unit 35 calculates paste processing for one path in step 161, sends the allocation history to the allocation history management unit 36 based on the calculation result, and generates an empty update trigger. It is sent to the table empty management unit 34 (step 162), a notification that the pasting process has been performed (pasting process notice) is sent to the band change counter 37 (step 163), and a table update trigger is sent to the link schedule table 25. (Step 164). Thus, the table empty management unit 34 updates the table empty information in step 165, and updates the link schedule table 25 in step 166. In the band change counter 37, the pasting counter is incremented in step 167, and it is determined in step 168 whether or not the counter value has reached the upper limit value Ra. If the upper limit value Ra has not been reached, the band change counter 37 returns to the state of waiting for the next pasting process notification. If the upper limit value Ra has been reached, the band change counter 37 sends an abort instruction to the increment pasting calculation unit 35 (step 169). In step 170, the increment paste calculation unit 35 determines whether or not there is an abort instruction, and if there is no abort instruction, the process returns to step 161 and executes the calculation for the next one-pass paste process. .

  After these processes, the process of step 131 shown in FIG. 19 is executed, and the scheduling process is completed.

  Next, a configuration for reducing the time until control reflection will be described.

  In an optical TDM network, each node needs to recognize a processing schedule such as a data transmission schedule and a switch route switching schedule for its normal operation. For this reason, a node schedule table (or processing schedule) is required. Must be distributed to each node in advance. In the scheduler described in Method 1 to Method 3, although the link schedule table 25 itself is only partially changed, the entire table is converted into the node schedule table 27 from the beginning after the entire link schedule table 25 is completed. Therefore, there is a potential problem that the time required from when the schedule is confirmed to when it is reflected in the node schedule table 27 and reflected in each node becomes long. Therefore, it is conceivable to derive the node schedule table 27 with a low load and to shorten the time from when the schedule is confirmed until it is reflected in the actual control at each node.

  The scheduler shown in FIG. 38 has the same configuration as that of the scheduler for method 1 shown in FIG. 17, but a table rewriting unit 38 is provided instead of the table conversion unit, and the reduction schedule stripping process is performed in the link schedule table 25. Every time there is a partial change by the unit 33 or the increment paste processing unit 35, the partial change is immediately reflected in the node schedule table 27 by the table rewriting unit 38. The table rewriting unit 38 receives the results of the stripping process and the pasting process from the decrease stripping processing unit 33 and the increase part pasting processing unit 35, respectively, and based on those results, the table rewriting unit 38 performs stripping processing or pasting processing. The node schedule table 27 is partially changed for the location to be changed. By providing such a table rewriting unit 38, it is not necessary to perform conversion from the beginning after the completion of the link schedule table 25 to generate the node schedule table 27, which is reflected in the control at each node from the schedule determination. Can be shortened.

  Here, it has been explained that the time until the control reflection is shortened for the scheduler used in method 1, but the scheduler used in method 2 or method 3 is also provided with the table rewriting unit described above instead of the table conversion unit. Thus, the time until the control is reflected can be shortened.

  Each of the scheduling methods described above is a method of scheduling based on the difference in traffic volume between the previous time and the current time. However, if scheduling that considers only the difference is continued, there is a deviation from the optimal scheduling solution. Events may occur. In particular, in Method 1 that relies only on the difference in the number of required time slots, it is considered that deviation from the optimal solution is likely to occur. Therefore, in order to deal with the deviation from the optimal solution, it is conceivable to periodically correct the optimal solution.

  FIG. 39 shows the configuration of the scheduler that suppresses the deviation from the optimal scheduling solution. In the illustrated scheduler 60, two types of schedulers 62 and 63 are switched by a selector 64, and the link schedule table 25 is changed according to the processing result of the scheduler selected by the selector 64. It has become. The scheduler 60 includes the AC traffic amount totaling unit 21, the table conversion unit 26, the node schedule table 27, the table transmission unit 28, the timer 29, and the allocation history management unit 36, as in the schedulers of the method 1 to the method 3 described above. Furthermore, a traffic tendency calculation unit 61 and a correction trigger unit 65 are also provided.

  The traffic tendency calculation unit 61 acquires information on the number of required time slots from the AC traffic volume totaling unit 21 and compares the information with the past required time slot number information, thereby increasing or decreasing traffic for each ground (path) and their changes. Calculate the trend of quantity.

  The scheduler 62 receives the calculation results on the traffic increase / decrease and the trend of the change amount from the traffic trend calculation unit 62, receives the past allocation history information from the allocation history management unit 36, and has conventionally known general scheduling. The optimal scheduling solution is derived according to the algorithm. The scheduler 62 outputs a trigger signal for notifying the fact to the correction trigger unit 65 when the optimum solution is derived.

  On the other hand, the scheduler 63 is a scheduler that performs a stripping process and a pasting process based on a difference according to the present invention. Specifically, the scheduler 63, for example, the traffic fluctuation calculation unit 31, the fluctuation difference table 32, the decrease stripping calculation unit 33, the table empty management unit 34, and the increase paste in the scheduler for the method 1 shown in FIG. A portion comprising the calculation unit 35, or an allocation difference calculation unit 41, an allocation difference table 42, an excess stripping calculation unit 43, a table empty management unit 34, and a shortage paste calculation in the scheduler for the method 2 shown in FIG. It is comprised by the part which consists of the part 44. Further, the scheduler 63 may operate based on the method 3.

  The correction trigger unit 65 updates the entire link schedule table 25 with the optimal scheduling solution output from the scheduler 62 only when a trigger signal is issued from the scheduler 62, and the stripping executed by the scheduler 63 otherwise. The selector 54 is switched so that the link schedule table 25 is partially changed by the taking process and the pasting process.

  The time required for the calculation for deriving the optimal scheduling solution is considerably longer than the time required for the calculation for performing the stripping process and the pasting process based on the difference. Therefore, in the scheduler 60 shown in FIG. 39, the scheduler 62 for deriving the optimal scheduling solution performs scheduling calculation with a period longer than the scheduling period by the stripping process and the pasting process, and scheduling by the stripping process and the pasting process. Correction for By using the scheduler 60, scheduling by the stripping process and the pasting process based on the difference makes it possible to shorten the time required for scheduling and execute resource allocation at a high speed, and to deviate from the optimal scheduling solution. Can be minimized.

10 link 11 node 12 host computer 20, 60, 61, 62 scheduler 21 AC traffic volume totaling unit 22 unallocated path table 23 path selector 24 table writing unit 25 link schedule table 26 table conversion unit 27 node schedule table 28 table transmission unit DESCRIPTION OF SYMBOLS 29 Timer 31 Traffic fluctuation | variation difference calculation part 32 Fluctuation difference table 33 Decrease stripping calculation part 34 Table empty management part 35 Increase sticking calculation part 36 Allocation history management part 37 Band change counter 38 Table rewriting part 41 Allocation difference calculation part 42 Allocation difference table 43 Excess stripping calculation unit 44 Insufficient pasting calculation unit 51 Comparator 52 Stripping calculation unit 53 Pasting calculation unit

Claims (7)

In a TDM network system comprising a plurality of nodes and a link connecting the nodes,
For each ground that is a combination of a communication source node and a destination node, a scheduler that assigns a time slot to the ground according to the amount of traffic in the ground,
The scheduler
A link schedule table showing the allocation status of time slots for each link and for each time slot position;
For each ground, a difference detection means for detecting a traffic amount difference between the previous scheduling execution and the current scheduling execution;
A stripping calculation means for performing stripping processing for deleting a time slot assigned to the ground from the link schedule table in response to a decrease in traffic volume of the ground for each ground,
In accordance with an increase in the traffic volume of the ground for each ground, a paste calculation means for performing a paste process for adding a time slot for the ground at an empty position in the link schedule table;
A TDM network system comprising: The difference detection means compares the required traffic slot total for generating the required time slot matrix by converting the required traffic volume of each ground into the number of time slots, and the current required time slot matrix and the previous required time slot matrix. And a traffic fluctuation difference calculation unit that stores it in the fluctuation difference table as fluctuation difference information,
The stripping calculation unit and the pasting calculation unit respectively perform the stripping process and the pasting process with reference to the variation difference table,
The TDM network system according to claim 1, wherein the stripping process is performed with priority over the pasting process. The difference detection means includes an AC traffic amount totaling unit that generates a required time slot matrix by converting the required traffic amount of each ground into the number of time slots, and an assigned time slot that indicates the number of time slots currently assigned to each ground. An allocation difference calculation unit that compares the matrix with the current required time slot matrix and stores it in the allocation difference table as allocation difference information,
The stripping calculation unit and the pasting calculation unit respectively perform the stripping process and the pasting process with reference to the allocation difference table,
2. The TDM network system according to claim 1, wherein the affixing process is prioritized over the ablation process, and the ablation process is executed when the affixing process cannot be completed. 3. The scheduler further comprises a comparator for determining a magnitude relationship between the number of grounds with increased traffic and the number of grounds with decreased traffic,
The difference detection means compares the required traffic slot total for generating the required time slot matrix by converting the required traffic volume of each ground into the number of time slots, and the current required time slot matrix and the previous required time slot matrix. Then, the traffic fluctuation difference calculation unit that stores the fluctuation difference information in the fluctuation difference table, the assigned time slot matrix indicating the number of time slots currently assigned to each ground, and the current required time slot matrix are compared and assigned. An allocation difference calculation unit that stores it in the allocation difference table as difference information,
The comparator selects either the variation difference table or the allocation difference table for the stripping calculation unit and the pasting calculation unit according to the magnitude relationship, and performs the stripping process and the pasting process. Let
When the stripping process and the pasting process are performed with reference to the variation difference table, the stripping process is executed in preference to the pasting process,
When the stripping process and the pasting process are performed with reference to the allocation difference table, the pasting process has priority over the stripping process, and the pasting process cannot be completed. The TDM network system according to claim 1, wherein the stripping process is executed. The scheduler further includes a bandwidth change counter that counts and manages the number of executions of the stripping process and the number of executions of the pasting process,
When the number of executions of the stripping process reaches a first upper limit, the execution of the stripping process is aborted, and when the number of executions of the pasting process reaches a second upper limit, the pasting process The TDM network system according to any one of claims 1 to 4, wherein execution of is terminated. In a TDM network system including a plurality of nodes and a link connecting the nodes, and assigning a time slot to the ground according to the amount of traffic in the ground for each ground that is a combination of a communication source node and a destination node In the scheduling method,
For each ground, detecting a difference in traffic between the previous scheduling execution and the current scheduling execution;
Stripping to delete the time slot assigned to the ground from the link schedule table indicating the allocation status of the time slot for each link and for each time slot position in response to a decrease in the traffic volume of the ground for each ground. The stage of processing,
Performing a pasting process of adding a time slot for the ground at an empty position in the link schedule table in response to an increase in traffic volume of the ground for each ground;
A scheduling method characterized by comprising: Provided in a TDM network system having a plurality of nodes and links connecting the nodes, a time slot is set for each ground according to the amount of traffic in the ground for each ground that is a combination of a communication source node and a destination node. Assign computers
For each ground, a difference detection means for detecting a traffic amount difference between the previous scheduling execution and the current scheduling execution,
Stripping to delete the time slot assigned to the ground from the link schedule table indicating the allocation status of the time slot for each link and for each time slot position in response to a decrease in the traffic volume of the ground for each ground. A stripping calculation means for executing processing, and a paste for adding a time slot for the ground at an empty position in the link schedule table in response to an increase in the traffic volume of the ground for each ground. Paste calculation means for executing the pasting process,
Program to function as.

Images